EP1236173A2 - Procedes et dispositifs pouvant identifier des modeles dans des systemes biologiques, et procedes d'utilisation - Google Patents

Procedes et dispositifs pouvant identifier des modeles dans des systemes biologiques, et procedes d'utilisation

Info

Publication number
EP1236173A2
EP1236173A2 EP00973988A EP00973988A EP1236173A2 EP 1236173 A2 EP1236173 A2 EP 1236173A2 EP 00973988 A EP00973988 A EP 00973988A EP 00973988 A EP00973988 A EP 00973988A EP 1236173 A2 EP1236173 A2 EP 1236173A2
Authority
EP
European Patent Office
Prior art keywords
training
data set
test
support vector
processing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
EP00973988A
Other languages
German (de)
English (en)
Inventor
Stephen D. Barnhill
Isabelle Guyon
Jason Weston
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Health Discovery Corp
Original Assignee
BIOWULF TECHNOLOGIES LLC
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from US09/568,301 external-priority patent/US6427141B1/en
Priority claimed from US09/578,011 external-priority patent/US6658395B1/en
Application filed by BIOWULF TECHNOLOGIES LLC filed Critical BIOWULF TECHNOLOGIES LLC
Priority to EP10185728A priority Critical patent/EP2357582A1/fr
Publication of EP1236173A2 publication Critical patent/EP1236173A2/fr
Withdrawn legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
    • G16B40/30—Unsupervised data analysis
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61P—SPECIFIC THERAPEUTIC ACTIVITY OF CHEMICAL COMPOUNDS OR MEDICINAL PREPARATIONS
    • A61P35/00—Antineoplastic agents
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B25/00—ICT specially adapted for hybridisation; ICT specially adapted for gene or protein expression
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
    • G16B40/20—Supervised data analysis

Definitions

  • the present invention relates to the use of learning machines to identify relevant patterns in biological systems such as genes, gene products, proteins, lipids, and combinations of the same. These patterns in biological systems can be used to diagnose and prognose abnormal physiological states. In addition, the patterns that can be detected using the present invention can be used to develop therapeutic agents.
  • oligonucleotide probes attached a solid base structure. Such devices are described in U.S. Patent Nos. 5,837,832 and 5,143,854, herein incorporated by reference in their entirety.
  • the oligonucleotide probes present on the chip can be used to determine whether a target nucleic acid has a nucleotide sequence identical to or different from a specific reference sequence.
  • the array of probes comprise probes that are complementary to the reference sequence as weU as probes that differ by one of more bases from the complementary probes.
  • the gene chips are capable of containing large arrays of oliogonucleotides on very small chips.
  • Methods for measuring hybridization intensity data to determine which probes are hybridizing are known in the art. Methods for detecting hybridization include fluorescent, radioactive, enzymatic, chemoluminescent, bioluminescent and other detection systems. Older, but still usable, methods such as gel electrophosesis and hybridization to gel blots or dot blots are also useful for determining genetic sequence information. Capture and detection systems for solution hybridization and in situ hybridization methods are also used for determining information about a genome.
  • proteomics is the study of the group of proteins encoded and regulated by a genome. This field represents a new focus on analyzing proteins, regulation of protein levels and the relationship to gene regulation and expression. Understanding the normal or pathological state of the proteome of a person or a population provides information for the prognosis or diagnosis of disease, development of drug or genetic treatments, or enzyme replacement therapies.
  • Current methods of studying the proteome involve 2-dimensional (2-D) gel electrophoresis of the proteins followed by analysis by mass spectrophotometry. A pattern of proteins at any particular time or stage in pathologenesis or treatment can be observed by 2-D gel electrophoresis.
  • the mass spectrophotometer is used to identify a protein isolated from the gel by identifying the amino acid sequence and comparing it to known sequence databases. Unfortunately, these methods require multiple steps to analyze a small portion of the proteome.
  • Genomic activation or expression does not always mean direct changes in protein production levels or activity.
  • Alternative processing of mRNA or post-transcriptional or post-translational regulatory mechanisms may cause the activity of one gene to result in multiple proteins, all of which are slightly different with different migration patterns and biological activities.
  • the human genome potentially contains 100,000 genes but the human proteome is beheved to be 50 to 100 times larger.
  • Knowledge discovery is the most desirable end product of data collection. Recent advancements in database technology have lead to an explosive growth in systems and methods for generating, collecting and storing vast amounts of data. While database technology enables efficient collection and storage of large data sets, the challenge of facilitating human comprehension of the information in this data is growing ever more difficult. With many existing techniques the problem has become unapproachable. Thus, there remains a need for a new generation of automated knowledge discovery tools. As a specific example, the Human Genome Project is populating a multi- gigabyte database describing the human genetic code. Before this mapping of the human genome is complete, the size of the database is expected to grow significantly. The vast amount of data in such a database overwhelms traditional tools for data analysis, such as spreadsheets and ad hoc queries.
  • Back-propagation neural networks are learning machines that may be trained to discover knowledge in a data set that may not be readily apparent to a human.
  • back-propagation neural network approaches that prevent neural networks from being well-controlled learning machines.
  • a significant drawback of back- propagation neural networks is that the empirical risk function may have many local minimums, a case that can easily obscure the optimal solution from discovery by this technique.
  • Standard optimization procedures employed by back-propagation neural networks may converge to an answer, but the neural network method cannot guarantee that even a localized minimum is attained much less the desired global minimum.
  • the quality of the solution obtained from a neural network depends on many factors. In particular the skill of the practitioner implementing the neural network determines the ultimate benefit, but even factors as seemingly benign as the random selection of initial weights can lead to poor results.
  • the convergence of the gradient based method used in neural network learning is inherently slow.
  • the sigmoid activation function has a scaling factor, which affects the quality of approximation. Possibly the largest limiting factor of neural networks as related to knowledge discovery is the "curse of dimensionality" associated with the disproportionate growth in required computational time and power for each additional feature or dimension in the training data.
  • a support vector machine maps input vectors into high dimensional feature space through non -linear mapping function, chosen a priori.
  • an optimal separating hyperplane is constructed.
  • the optimal hyperplane is then used to determine things such as class separations, regression fit, or accuracy in density estimation.
  • the dimensionally of the feature space may be huge.
  • a fourth degree polynomial mapping function causes a 200 dimensional input space to be mapped into a 1.6 billionth dimensional feature space.
  • Patent applications directed to support vector machines include, U.S. Patent Application Nos. 09/303,386; 09/303,387; 09/303,389; 09/305.345; all filed May 1, 1999; and U.S. Patent Application No. 09/568,301, filed May 9, 2000; and U.S. Patent Application No. 09/578,011, filed May 24, 2000 and also claims the benefit of U.S. Provisional Patent Application No. 60/161,806, filed October 27, 1999; of U.S. Provisional Patent
  • the optimal hyperplane can be constructed from a small number of support vectors relative to the training set size, the generalization ability will be high, even in infinite dimensional space.
  • the data generated from genomic and proteomic tests can be analyzed from many different viewpoints.
  • the literature shows simple approaches such as studies of gene clusters discovered by unsupervised learning techniques (Alon, 1999). Clustering is often also done along the other dimension of the data. For example, each experiment may correspond to one patient carrying or not carrying a specific disease
  • Support vector machines provide a desirable solution for the problem of discovering knowledge from vast amounts of input data.
  • the ability of a support vector machine to discover knowledge from a data set is limited in proportion to the information included within the training data set.
  • a system and method for pre-processing data so as to augment the training data to maximize the knowledge discovery by the support vector machine.
  • the raw output from a support vector machine may not fully disclose the knowledge in the most readily interpretable form.
  • a system and method for post-processing data output from a support vector machine in order to maximize the value of the information delivered for human or further automated processing.
  • the ability of a support vector machine to discover knowledge from data is limited by the selection of a kernel. Accordingly, there remains a need for an improved system and method for selecting and/or creating a desired kernel for a support vector machine.
  • the present invention comprises systems and methods for enhancing knowledge discovered from data using a learning machine in general and a support vector machine in particular.
  • the present invention comprises methods of using a learning machine for diagnosing and prognosing changes in biological systems such as diseases.
  • the specific relationships discovered are used to diagnose and prognose diseases, and methods of detecting and treating such diseases are applied to the biological system.
  • One embodiment of the present invention comprises preprocessing a training data set in order to allow the most advantageous application of the learning machine.
  • Each training data point comprises a vector having one or more coordinates.
  • Preprocessing the training data set may comprise identifying missing or erroneous data points and taking appropriate steps to correct the flawed data or as appropriate remove the observation or the entire field from the scope of the problem.
  • Pre-processing the training data set may also comprise adding dimensionality to each training data point by adding one or more new coordinates to the vector.
  • the new coordinates added to the vector may be derived by applying a transformation to one or more of the original coordinates.
  • the transformation may be based on expert knowledge, or may be computationally derived. In a situation where the training data set comprises a continuous variable, the transformation may comprise optimally categorizing the continuous variable of the training data set.
  • the support vector machine is trained using the pre- processed training data set.
  • the additional representations of the training data provided by the preprocessing may enhance the learning machine's ability to discover knowledge therefrom.
  • the greater the dimensionality of the training set the higher the quality of the generalizations that may be derived therefrom.
  • the training output may be post-processed by optimally categorizing the training output to derive categorizations from the continuous variable.
  • a test data set is pre-processed in the same manner as was the training data set. Then, the trained learning machine is tested using the pre-processed test data set.
  • a test output of the trained learning machine may be post-processing to determine if the test output is an optimal solution. Post-processing the test output may comprise interpreting the test output into a format that may be compared with the test data set. Alternative postprocessing steps may enhance the human interpretability or suitability for additional processing of the output data.
  • the present invention also provides for the selection of at least one kernel prior to training the support vector machine.
  • the selection of a kernel may be based on prior knowledge of the specific problem being addressed or analysis of the properties of any available data to be used with the learning machine and is typically dependant on the nature of the knowledge to be discovered from the data.
  • an iterative process comparing postprocessed training outputs or test outputs can be applied to make a determination as to which configuration provides the optimal solution. If the test output is not the optimal solution, the selection of the kernel may be adjusted and the support vector machine may be retrained and retested.
  • a live data set may be collected and pre-processed in the same manner as was the training data set.
  • the pre-processed live data set is input into the learning machine for processing.
  • the live output of the learning machine may then be post-processed by interpreting the live output into a computationally derived alphanumeric classifier or other form suitable to further utilization of the SVM derived answer.
  • a system enhancing knowledge discovered from data using a support vector machine.
  • the exemplary system comprises a storage device for storing a training data set and a test data set, and a processor for executing a support vector machine.
  • the processor is also operable for collecting the training data set from the database, pre-processing the training data set to enhance each of a plurality of training data points, training the support vector machine using the pre- processed training data set, collecting the test data set from the database, pre-processing the test data set in the same manner as was the training data set, testing the trained support vector machine using the pre-processed test data set, and in response to receiving the test output of the trained support vector machine, post-processing the test output to determine if the test output is an optimal solution.
  • the exemplary system may also comprise a communications device for receiving the test data set and the training data set from a remote source.
  • the processor may be operable to store the training data set in the storage device prior pre-processing of the training data set and to store the test data set in the storage device prior pre-processing of the test data set.
  • the exemplary system may also comprise a display device for displaying the post- processed test data.
  • the processor of the exemplary system may further be operable for performing each additional function described above.
  • the communications device may be further operable to send a computationally derived alphanumeric classifier or other SVM-based raw or post-processed output data to a remote source.
  • a system and method are provided for enhancing knowledge discovery from data using multiple learning machines in general and multiple support vector machines in particular.
  • Training data for a learning machine is pre- processed in order to add meaning thereto.
  • Pre-processing data may involve transforming the data points and/or expanding the data points.
  • the learning machine is provided with a greater amount of information for processing.
  • support vector machines the greater the amount of information that is processed, the better generalizations about the data that may be derived.
  • Multiple support vector machines, each comprising distinct kernels are trained with the pre-processed training data and are tested with test data that is pre-processed in the same manner. The test outputs from multiple support vector machines are compared in order to determine which of the test outputs if any represents an optimal solution.
  • Selection of one or more kernels may be adjusted and one or more support vector machines may be retrained and retested.
  • live data is pre-processed and input into the support vector machine comprising the kernel that produced the optimal solution.
  • the live output from the learning machine may then be post-processed into a computationally derived alphanumeric classifier for interpretation by a human or computer automated process.
  • a data set representing a continuous variable comprises data points that each comprise a sample from the continuous variable and a class identifier.
  • a number of distinct class identifiers within the data set is determined and a number of candidate bins is determined based on the range of the samples and a level of precision of the samples within the data set.
  • Each candidate bin represents a sub-range of the samples.
  • the entropy of the data points falling within the candidate bin is calculated.
  • a cutoff point in the range of samples is defined to be at the boundary of the last candidate bin in the sequence of candidate bins.
  • the collective entropy for different combinations of sequential candidate bins may be calculated.
  • the number of defined cutoff points may be adjusted in order to determine the optimal number of cutoff points, which is based on a calculation of minimal entropy.
  • the exemplary system and method for optimally categorizing a continuous variable may be used for pre-processing data to be input into a learning machine and for post-processing output of a learning machine.
  • a system and method are provided for for enhancing knowledge discovery from data using a learning machine in general and a support vector machine in particular in a distributed network environment.
  • a customer may transmit training data, test data and Uve data to a vendor's server from a remote source, via a distributed network.
  • the customer may also transmit to the server identification information such as a user name, a password and a financial account identifier.
  • the training data, test data and live data may be stored in a storage device.
  • Training data may then be pre-processed in order to add meaning thereto.
  • Pre- processing data may involve transforming the data points and/or expanding the data points. By adding meaning to the data, the learning machine is provided with a greater amount of information for processing.
  • the learning machine is therefore trained with the pre-processed training data and is tested with test data that is pre- processed in the same manner.
  • the test output from the learning machine is post- processed in order to determine if the knowledge discovered from the test data is desirable.
  • Post-processing involves interpreting the test output into a format that may be compared with the test data. Live data is pre-processed and input into the trained and tested learning machine.
  • the Uve output from the learning machine may then be post- processed into a computationally derived alphanumerical classifier for interpretation by a human or computer automated process.
  • the server Prior to transmitting the alpha numerical classifier to the customer via the distributed network, the server is operable to communicate with a financial institution for the purpose of receiving funds from a financial account of the customer identified by the financial account identifier.
  • FIG. 1 is a flowchart illustrating an exemplary general method for increasing knowledge that may be discovered from data using a learning machine.
  • FIG. 2 is a flowchart illustrating an exemplary method for increasing knowledge that may be discovered from data using a support vector machine.
  • FIG. 3 is a flowchart illustrating an exemplary optimal categorization method that may be used in a stand-alone configuration or in conjunction with a learning machine for pre-processing or post-processing techniques in accordance with an exemplary embodiment of the present invention.
  • FIG. 4 illustrates an exemplary unexpanded data set that may be input into a support vector machine.
  • FIG. 5 illustrates an exemplary post-processed output generated by a support vector machine using the data set of FIG. 4.
  • FIG. 6 illustrates an exemplary expanded data set that may be input into a support vector machine based on the data set of FIG. 4.
  • FIG. 7 illustrates an exemplary post-processed output generated by a support vector machine using the data set of FIG. 6.
  • FIG. 8 illustrates exemplary input and output for a standalone application of the optimal categorization method of FIG. 3.
  • FIG. 9 is a comparison of exemplary post-processed output from a first support vector machine comprising a linear kernel and a second support vector machine comprising a polynomial kernel.
  • FIG. 10 is a functional block diagram illustrating an exemplary operating environment for an exemplary embodiment of the present invention.
  • FIG. 11 is a functional block diagram illustrating an alternate exemplary operating environment for an alternate embodiment of the present invention.
  • FIG. 12 is a functional block diagram illustrating an exemplary network operating environment for implementation of a further alternate embodiment of the present invention.
  • FIG. 13 graphically illustrates use of a linear discriminant classifier.
  • FIG. 14 shows graphs of the results of using RFE with information similar to Example 2.
  • FIG. 15 shows the distribution of gene expression values across tissue samples for two genes.
  • FIG. 16 shows the distribution of gene expression values across genes for aU tissue samples.
  • FIG. 17 shows the data matrices representing gene expression values from microarray data for colon cancer.
  • FIG. 18 shows the results of RFE after preprocessing.
  • FIG. 19 shows a graphical comparison with the present invention and the methods of Golub.
  • FIG. 20 shows the correlation between the best 32 genes and all other genes.
  • FIG. 21 shows the results of RFE when training on 100 dense QT_clust clusters.
  • FIG. 22 shows the top 8 QT_clust clusters chosen by SVM RFE.
  • FIG. 23 shows the QT_clust top gene scatter plot
  • FIG. 24 shows supervised clustering.
  • FIG. 25 shows the results of SVM RFE when training on the entire data set.
  • FIG. 26 shows the results of Golub's method when training on the entire data set.
  • FIG. 27 shows the weighting coefficients of the support vectors.
  • FIG. 28 shows the top ranked genes discovered by SVM RFE in order of increasing importance from left to right.
  • FIG. 29 shows the 7 top ranked genes discovered by Golub's methods in order of increasing improtance from left to right.
  • FIG. 30 shows a comparison of feature (gene) selection methods for colon cancer data using different methods.
  • FIG. 31 shows the metrics of classifier quality.
  • the triangle and circle curves represent example distributions of two classes: class 1 (negative class) and class 2 (positive class).
  • FIG. 32A and 32B show the performance comparison between SVMs and the baseline method for leukemia data.
  • FIG. 33 shows the best set of 16 genes for the leukemia data.
  • FIG. 34 shows the selection of an optimum number of genes for leukemia data.
  • FIG. 35 shows the selection of an optimum number of genes for colon cancer data.
  • FIG. 36 is a functional block diagram illustrating a hierarchical system of multiple support vector machine.
  • the present invention provides methods, systems and devices for discovering knowledge from data using learning machines.
  • the present invention is directed to methods, systems and devices for knowledge discovery from data using learning machines that are provided information regarding changes in biological systems. More particularly, the present invention comprises methods of use of such knowledge for diagnosing and prognosing changes in biological systems such as diseases. Additionally, the present invention comprises methods, compositions and devices for applying such knowledge to the testing and treating of individuals with changes in their individual biological systems.
  • biological data means any data derived from measuring biological conditions of human, animals or other biological organisms including microorganisms, viruses, plants and other living organisms. The measurements may be made by any tests, assays or observations that are known to physicians, scientists, diagnosticians, or the like. Biological data may include, but is not limited to, clinical tests and observations, physical and chemical measurements, genomic determinations, proteomic determinations, drug levels, hormonal and immunological tests, neurochemical or neurophysical measurements, mineral and vitamin level determinations, genetic and familial histories, and other determinations that may give insight into the state of the individual or individuals that are undergoing testing.
  • data is used interchangeably with “biological data”.
  • learning machines comprise algorithms that may be trained to generalize using data with known outcomes. Trained learning machine algorithms may then be applied to cases of unknown outcome for prediction.
  • a learning machine may be trained to recognize patterns in data, estimate regression in data or estimate probability density within data.
  • Learning machines may be trained to solve a wide variety of problems as known to those of ordinary skill in the art.
  • a trained learning machine may optionally be tested using test data to ensure that its output is validated within an acceptable margin of error.
  • live data may be input therein. The live output of a learning machine comprises knowledge discovered from all of the training data as applied to the live data.
  • the present invention comprises methods, systems and devices for analyzing patterns found in biological data, data such as that generated by examination of genes, transcriptional and translational products and proteins.
  • Genomic information can be found in patterns generated by hybridization reactions of genomic fragments and complementary nucleic acids or interacting proteins.
  • One of the most recent tools for investigating such genomic or nucleic acid interactions is the DNA gene chip or microarray.
  • the microarray allows for the processing of thousands of nucleic interactions. DNA microarrays enable researchers to screen thousands of genes in one experiment. For example, the microarray could contain 2400 genes on a small glass slide and can be used to determine the presence of DNA or RNA in the sample.
  • microarray tests can be used in basic research and biomedical research including tumor biology, neurosciences, signal transduction, transcription regulation, and cytokine and receptor studies. Additionally, there are applications for pharmaceutical drug discovery, target identification, lead optimization, pharmacokinetics, pharmacogenomics and diagnostics.
  • the market for microarray technology was approximately $98 million in 1999 and the amount of data generated and stored in databases developed from multiple microarray tests is enormous.
  • the present invention provides for methods, systems and devices that can use the data generated in such microarray and nucleic acid chip tests for the diagnosis and prognosis of diseases and for the development of therapeutic agents to treat such diseases.
  • the present invention also comprises devices comprising microarrays with specific sequence identifying probes that can be used to diagnose or prognose the specific change in the biological system.
  • the learning machine of the present invention Once the learning machine of the present invention has identified specific relationships among the data that are capable of diagnosing or prognosing a change in a biological system, specific devices then incorporate tests for those specific relationships. For example, the learning machine of the present invention identifies specific genes that are related to the presence or future occurrence of a change in a biological system, such as the presence or appearance of a tumor. Knowing the sequence of these genes allows for the making of a specific treating device for those identified genes. For example, a nucleic acid chip, comprising
  • DNA, RNA or specific binding proteins, or any such combination, that specifically binds to specifically identified genes is used to easily identify individuals having a particular tumor or the likelihood of developing the tumor.
  • specific proteins either identified by the learning machine or that are associated with the genes identified by the learning machine, can be tested for using serological tests directed to specifically detecting the identified proteins, gene products or antibodies or antibody fragments directed to the proteins or gene products.
  • Such tests include, but are not limited to, antibody microarrays on chips, Western blotting tests, ELISA, and other tests known in the art wherein binding between specific binding partners is used for detection of one of the partners.
  • the present invention comprises methods and compositions for treating the conditions resulting from changes in biological systems or for treating the biological system to alter the biological system to prevent or enhance specific conditions.
  • the individual can be treated with anti-tumor medications such as chemotherapeutic compositions.
  • the diagnosis of an individual includes the predisposition or prognosis of tumor development, the individual may be treated prophalactically with chemotherapeutic compositions to prevent the occurrence of the tumor.
  • specific genes are identified with the occurrence of tumors, the individual may be treated with specific antisense or other gene therapy methods to suppress the expression of such genes.
  • Proteomic investigations provide for methods of determining the proteins involved in normal and pathological states.
  • Current methods of determining the proteome of a person or a population at any particular time or stage comprise the use of gel electrophoresis to separate the proteins in a sample.
  • 2-D gel electrophoresis is used to separate the proteins more completely.
  • the sample may be preprocessed to remove known proteins.
  • the proteins may be labeled, for example, with fluorescent dyes, to aid in the determination of the patterns generated by the selected proteome. Patterns of separated proteins can be analyzed using the learning machines of the present invention. Capturing the gel image can be accomplished by image technology methods known in the art such as densitometry, CCD camera and laser scanning and storage phosphor instruments. Analysis of the gels reveals patterns in the proteome that are important in diagnosis and prognosis of pathological states and shows changes in relation to therapeutic interventions.
  • Such a technology comprises translating mRNA in vitro and covalently attaching the translated protein to the mRNA.
  • the sequence of the mRNA and thus the protein is then determined using amplification methods such as PCR.
  • Libraries containing 10 14 to 10 15 members can be established from this data. These libraries can be used to determine peptides that bind receptors or antibody libraries can be developed that contain antibodies that avidly bind their targets.
  • Libraries called protein domain libraries can be created from cellular mRNA where the entire proteins are not translated, but fragments are sequenced. These libraries can be used to determine protein function.
  • Genomic libraries can be searched for open reading frames (ORFS) or ESTs (expressed sequence tags) of interest and from the sequence, peptides are synthesized. Peptides for different genes are placed in 96 well trays for selection of antibodies from phage libraries. The antibodies are then used to locate the protein relating to the original ORFs or ESTs in sections of normal and diseased tissue.
  • ORFS open reading frames
  • ESTs expressed sequence tags
  • the present invention can be used to analyze biological data generated at multiple stages of investigation into biological functions, and further, to integrate the different kinds of data for novel diagnostic and prognostic determinations.
  • biological data obtained from clinical case information such as diagnostic test data, family or genetic histories, prior or current medical treatments, and the clinical outcomes of such activities, can be utilized in the methods, systems and devices of the present invention.
  • clinical samples such as diseased tissues or fluids, and normal tissues and fluids, and cell separations can provide biological data that can be utilized by the current invention.
  • Proteomic determinations such as 2-D gel, mass spectrophotometry and antibody screening can be used to establish databases that can be utilized by the present invention.
  • Genomic databases can also be used alone or in combination with the above-described data and databases by the present invention to provide comprehensive diagnosis, prognosis or predictive capabilities to the user of the present invention.
  • a first aspect of the present invention seeks to enhance knowledge discovery by optionally pre-processing data prior to using the data to train a learning machine and/or optionally post-processing the output from a learning machine.
  • preprocessing data comprises reformatting or augmenting the data in order to allow the learning machine to be applied most advantageously.
  • post-processing involves interpreting the output of a learning machine in order to discover meaningful characteristics thereof. The meaningful characteristics to be ascertained from the output may be problem or data specific.
  • Post-processing involves interpreting the output into a form that is comprehendible by a human or one that is comprehendible by a computer. Exemplary embodiments of the present invention wiU hereinafter be described with reference to the drawing, in which like numerals indicate like elements throughout the several figures.
  • step 102 is a flowchart illustrating a general method 100 for enhancing knowledge discovery using learning machines.
  • the method 100 begins at starting block 101 and progresses to step 102 where a specific problem is formalized for application of knowledge discovery through machine learning. Particularly important is a proper formulation of the desired output of the learning machine. For instance, in predicting future performance of an individual equity instrument, or a market index, a learning machine is likely to achieve better performance when predicting the expected future change rather than predicting the future price level. The future price expectation can later be derived in a post-processing step as will be discussed later in this specification.
  • step 103 addresses training data collection.
  • Training data comprises a set of data points having known characteristics. Training data may be collected from one or more local and/or remote sources. The collection of training data may be accomplished manually or by way of an automated process, such as known electronic data transfer methods. Accordingly, an exemplary embodiment of the present invention may be implemented in a networked computer environment.
  • FIGS. 10-12 Exemplary operating environments for implementing various embodiments of the present invention will be described in detail with respect to FIGS. 10-12.
  • the collected training data is optionally pre-processed in order to allow the learning machine to be applied most advantageously toward extraction of the knowledge inherent to the training data.
  • the training data can optionally be expanded through transformations, combinations or manipulation of individual or multiple measures within the records of the training data.
  • expanding data is meant to refer to altering the dimensionality of the input data by changing the number of observations available to determine each input point (alternatively, this could be described as adding or deleting columns within a database table).
  • a data point may comprise the coordinates (1,4,9).
  • An expanded version of this data point may result in the coordinates (1,1,4,2,9,3).
  • the coordinates added to the expanded data point are based on a square-root transformation of the original coordinates.
  • this expanded data point provides a varied representation of the input data that is potentially more meaningful for knowledge discovery by a learning machine. Data expansion in this sense affords opportunities for learning machines to discover knowledge not readily apparent in the unexpanded training data.
  • Expanding data may comprise applying any type of meaningful transformation to the data and adding those transformations to the original data.
  • the criteria for determining whether a transformation is meaningful may depend on the input data itself and/or the type of knowledge that is sought from the data.
  • Illustrative types of data transformations include: addition of expert information; labeling; binary conversion; sine, cosine, tangent, cotangent, and other trigonometric transformation; clustering; scaling; probabilistic and statistical analysis; significance testing; strength testing; searching for 2-D regularities; Hidden Markov Modeling; identification of equivalence relations; application of contingency tables; application of graph theory principles; creation of vector maps; addition, subtraction, multiplication, division, application of polynomial equations and other algebraic transformations; identification of proportionality; determination of discriminatory power; etc.
  • transformations include: association with known standard medical reference ranges; physiologic truncation; physiologic combinations; biochemical combinations; application of heuristic rules; diagnostic criteria determinations; clinical weighting systems; diagnostic transformations; clinical transformations; application of expert knowledge; labeling techniques; application of other domain knowledge; Bayesian network knowledge; etc.
  • a data point may comprise the coordinate (A, B, C).
  • a transformed version of this data point may result in the coordinates (1, 2, 3), where the coordinate "1" has some known relationship with the coordinate "A,” the coordinate "2” has some known relationship with the coordinate "B,” and the coordinate "3” has some known relationship with the coordinate "C.”
  • a transformation from letters to numbers may be required, for example, if letters are not understood by a learning machine.
  • Other types of transformations are possible without adding dimensionality to the data points, even with respect to data that is originally in numeric form.
  • pre-processing data to add meaning thereto may involve analyzing incomplete, corrupted or otherwise "dirty" data.
  • a learning machine cannot process "dirty" data in a meaningful manner.
  • a pre-processing step may involve cleaning up a data set in order to remove, repair or replace dirty data points.
  • the exemplary method 100 continues at step 106, where the learning machine is trained using the pre-processed data.
  • a learning machine is trained by adjusting its operating parameters until a desirable training output is achieved. The determination of whether a training output is desirable may be accomplished either manually or automatically by comparing the training output to the known characteristics of the training data. A learning machine is considered to be trained when its training output is within a predetermined error threshold from the known characteristics of the training data. In certain situations, it may be desirable, if not necessary, to post-process the training output of the learning machine at step 107. As mentioned, post-processing the output of a learning machine involves interpreting the output into a meaningful form.
  • test data is optionally collected in preparation for testing the trained learning machine.
  • Test data may be collected from one or more local and/or remote sources.
  • test data and training data may be collected from the same source(s) at the same time.
  • test data and training data sets can be divided out of a common data set and stored in a local storage medium for use as different input data sets for a learning machine.
  • any test data used must be pre-processed at step 110 in the same manner as was the training data.
  • a proper test of the learning may only be accomplished by using testing data of the same format as the training data.
  • the learning machine is tested using the pre-processed test data, if any.
  • the test output of the learning machine is optionally post-processed at step 114 in order to determine if the results are desirable.
  • the post processing step involves interpreting the test output into a meaningful form.
  • the meaningful form may be one that is comprehendible by a human or one that is comprehendible by a computer.
  • the test output must be post-processed into a form which may be compared to the test data to determine whether the results were desirable. Examples of post- processing steps include but are not limited of the following: optimal categorization determinations, scaling techniques (linear and non-linear), transformations (linear and non-linear), and probability estimations.
  • the method 100 ends at step 116.
  • FIG. 2 is a flow chart illustrating an exemplary method 200 for enhancing knowledge that may be discovered from data using a specific type of learning machine known as a support vector machine (SVM).
  • SVM implements a specialized algorithm for providing generalization when estimating a multi-dimensional function from a limited collection of data.
  • An SVM may be particularly useful in solving dependency estimation problems. More specifically, an SVM may be used accurately in estimating indicator functions (e.g. pattern recognition problems) and real-valued functions (e.g. function approximation problems, regression estimation problems, density estimation problems, and solving inverse problems).
  • indicator functions e.g. pattern recognition problems
  • real-valued functions e.g. function approximation problems, regression estimation problems, density estimation problems, and solving inverse problems.
  • the concepts underlying the SVM are explained in detail in a book by Vladimir N. Vapnikv, entitled Statistical Learning Theory (John Wiley & Sons, Inc. 1998), which is herein incorporated by reference in its entirety. Accordingly, a
  • a pattern recognition system using support vectors was disclosed in U.S. Patent No. 5,649,068, which is herein incorporated in its entirety.
  • a method is described in the patent wherein the dual representation mathematical principle was used for the design of decision systems. This principle permits some decision functions that are weighted sums of predefined functions to be represented as memory-based decision function. Using this principle, a memory-based decision system with optimum margin was designed wherein weights and prototypes of training patterns of a memory-based decision function were determined such that the corresponding dual decision function satisfies the criterion of margin optimality.
  • the exemplary method 200 begins at starting block 201 and advances to step
  • training data may be collected from one or more local and/or remote sources, through a manual or automated process.
  • the training data is optionally pre-processed.
  • pre-processing data comprises enhancing meaning within the training data by cleaning the data, transforming the data and/or expanding the data.
  • SVMs are capable of processing input data having extremely large dimensionality. In fact, the larger the dimensionality of the input data, the better generalizations an SVM is able to calculate. Though merely increasing the dimensionality of the input space through preprocessing does not guarantee better generalization with an SVM.
  • a kernel is selected for the SVM.
  • different kernels wiU cause an SVM to produce varying degrees of quality in the output for a given set of input data. Therefore, the selection of an appropriate kernel may be essential to the desired quality of the output of the SVM.
  • a kernel may be chosen based on prior performance knowledge.
  • exemplary kernels include polynomial kernels, radial basis function kernels, linear kernels, etc.
  • a customized kernel may be created that is specific to a particular problem or type of data set.
  • the multiple SVMs may be trained and tested simultaneously, each using a different kernel. The quality of the outputs for each simultaneously trained and tested SVM may be compared using a variety of selectable or weighted metrics (see step 222) to determine the most desirable kernel.
  • the pre-processed training data is input into the SVM.
  • the SVM is trained using the pre-processed training data to generate an optimal hyperplane.
  • the training output of the SVM may then be post-processed at step 211.
  • post-processing of training output may be desirable, or even necessary, at this point in order to properly calculate ranges or categories for the output.
  • test data is collected similarly to previous descriptions of data collection.
  • the test data is pre-processed at step 214 in the same manner as was the training data above.
  • the pre-processed test data is input into the SVM for processing in order to determine whether the SVM was trained in a desirable manner.
  • the test output is received from the SVM at step 218 and is optionally post-processed at step 220.
  • an SVM is operable to ascertain an output having a global minimum error.
  • output results of an SVM for a given data set will typically vary in relation to the selection of a kernel. Therefore, there are in fact multiple global minimums that may be ascertained by an SVM for a given set of data.
  • the term "optimal minimum” or "optimal solution” refers to a selected global minimum that is considered to be optimal (e.g. the optimal solution for a given set of problem specific, pre-established criteria) when compared to other global minimums ascertained by an SVM.
  • determining whether the optimal minimum has been ascertained may involve comparing the output of an SVM with a historical or predetermined value.
  • a predetermined value may be dependant on the test data set. For example, in the context of a pattern recognition problem where data points are classified by an SVM as either having a certain characteristic or not having the characteristic, a global minimum error of 50% would not be optimal. In this example, a global minimum of 50% is no better than the result that would be achieved by chance.
  • the outputs for each SVM may be compared with each other SVM's outputs to determine the practical optimal solution for that particular set of kernels.
  • the determination of whether an optimal solution has been ascertained may be performed manually or through an automated comparison process.
  • step 224 the kernel selection is adjusted. Adjustment of the kernel selection may comprise selecting one or more new kernels or adjusting kernel parameters. Furthermore, in the case where multiple SVMs were trained and tested simultaneously, selected kernels may be replaced or modified while other kernels may be re-used for control purposes.
  • the method 200 is repeated from step 208, where the previously pre-processed training data is input into the SVM for training purposes.
  • step 226 live data is collected similarly as described above. The desired output characteristics that were known with respect to the training data and the test data are not known with respect to the live data.
  • the live data is pre-processed in the same manner as was the training data and the test data.
  • the live pre-processed data is input into the SVM for processing.
  • the live output of the SVM is received at step 232 and is post-processed at step 234.
  • post-processing comprises converting the output of the SVM into a computationally derived alpha-numerical classifier, for interpretation by a human or computer.
  • the alphanumerical classifier comprises a single value that is easily comprehended by the human or computer.
  • the method 200 ends at step 236.
  • FIG. 3 is a flow chart illustrating an exemplary optimal categorization method
  • the exemplary optimal categorization method 300 may be used for pre-processing data or post-processing output from a learning machine in accordance with an exemplary embodiment of the present invention. Additionally, as wiU be described below, the exemplary optimal categorization method may be used as a stand-alone categorization technique, independent from learning machines. The exemplary optimal categorization method 300 begins at starting block
  • step 302 an input data set is received.
  • the input data set comprises a sequence of data samples from a continuous variable.
  • the data samples fall within two or more classification categories.
  • step 304 the bin and class-tracking variables are initialized.
  • bin variables relate to resolution and class-tracking variables relate to the number of classifications within the data set. Determining the values for initialization of the bin and class-tracking variables may be performed manually or through an automated process, such as a computer program from analyzing the input data set.
  • the data entropy for each bin is calculated. Entropy is a mathematical quantity that measures the uncertainty of a random distribution. In the exemplary method 300, entropy is used to gauge the gradations of the input variable so that maximum classification capabihty is achieved.
  • the method 300 produces a series of "cuts" on the continuous variable, such that the continuous variable may be divided into discrete categories.
  • the cuts selected by the exemplary method 300 are optimal in the sense that the average entropy of each resulting discrete category is minimized.
  • a determination is made as to whether all cuts have been placed within input data set comprising the continuous variable. If all cuts have not been placed, sequential bin combinations are tested for cutoff determination at step 310. From step 310, the exemplary method 300 loops back through step 306 and returns to step 308 where it is again determined whether all cuts have been placed within input data set comprising the continuous variable.
  • step 309 When aU cuts have been placed, the entropy for the entire system is evaluated at step 309 and compared to previous results from testing more or fewer cuts. If it cannot be concluded that a minimum entropy state has been determined, then other possible cut selections must be evaluated and the method proceeds to step 311. From step 311 a heretofore untested selection for number of cuts is chosen and the above process is repeated from step 304. When either the limits of the resolution determined by the bin width has been tested or the convergence to a minimum solution has been identified, the optimal classification criteria is output at step 312 and the exemplary optimal categorization method 300 ends at step 314.
  • the optimal categorization method 300 takes advantage of dynamic programming techniques.
  • dynamic programming techniques may be used to significantly improve the efficiency of solving certain complex problems through carefully structuring an algorithm to reduce redundant calculations.
  • the straightforward approach of exhaustively searching through all possible cuts in the continuous variable data would result in an algorithm of exponential complexity and would render the problem intractable for even moderate sized inputs.
  • the target function in this problem the average entropy, the problem may be divide into a series of sub-problems.
  • algorithmic sub-structures for solving each sub-problem and storing the solutions of the sub-problems, a great amount of redundant computation may be identified and avoided.
  • the exemplary optimal categorization method 300 may be implemented as an algorithm having a polynomial complexity, which may be used to solve large sized problems.
  • the exemplary optimal categorization method 300 may be used in pre-processing data and/or post-processing the output of a learning machine. For example, as a pre-processing transformation step, the exemplary optimal categorization method 300 may be used to extract classification information from raw data. As a post-processing technique, the exemplary optimal range categorization method may be used to determine the optimal cut-off values for markers objectively based on data, rather than relying on ad hoc approaches. As should be apparent, the exemplary optimal categorization method 300 has applications in pattern recognition, classification, regression problems, etc. The exemplary optimal categorization method 300 may also be used as a stand-alone categorization technique, independent from
  • FIG. 4 illustrates an exemplary unexpanded data set 400 that may be used as input for a support vector machine.
  • This data set 400 is referred to as "unexpanded" because no additional information has been added thereto.
  • the unexpanded data set comprises a training data set 402 and a test data set 404.
  • Both the unexpanded training data set 402 and the unexpanded test data set 404 comprise data points, such as exemplary data point 406, relating to historical clinical data from sampled medical patients.
  • the data set 400 may be used to train an SVM to determine whether a breast cancer patient will experience a recurrence or not.
  • Each data point includes five input coordinates, or dimensions, and an output classification shown as 406a-f which represent medical data collected for each patient.
  • the first coordinate 406a represents “Age”
  • the second coordinate 406b represents “Estrogen Receptor Level”
  • the third coordinate 406c represents “Progesterone Receptor Level”
  • the fourth coordinate 406d represents "Total Lymph
  • the fifth coordinate 406e represents “Positive (Cancerous) Lymph Nodes Extracted”
  • the output classification 406f represents the "Recurrence Classification.”
  • the important known characteristic of the data 400 is the output classification 406f (Recurrence Classification), which, in this example, indicates whether the sampled medical patient responded to treatment favorably without recurrence of cancer ("- 1 ") or responded to treatment negatively with recurrence of cancer (" 1 ”) .
  • This known characteristic will be used for learning while processing the training data in the SVM, will be used in an evaluative fashion after the test data is input into the SVM thus creating a "blind” test, and will obviously be unknown in the Uve data of current medical patients.
  • FIG. 5 illustrates an exemplary test output 502 from an SVM trained with the unexpanded training data set 402 and tested with the unexpanded data set 404 shown in FIG. 4.
  • the test output 502 has been post-processed to be comprehensible by a human or computer. As indicated, the test output 502 shows that 24 total samples (data points) were examined by the SVM and that the SVM incorrectly identified four of eight positive samples (50%) and incorrectly identified 6 of sixteen negative samples (37.5%).
  • FIG. 6 illustrates an exemplary expanded data set 600 that may be used as input for a support vector machine.
  • This data set 600 is referred to as "expanded" because additional information has been added thereto.
  • the expanded data set 600 is identical to the unexpanded data set 400 shown in FIG. 4.
  • the additional information supplied to the expanded data set has been supplied using the exemplary optimal range categorization method 300 described with reference to FIG. 3.
  • the expanded data set comprises a training data set 602 and a test data set 604.
  • Both the expanded training data set 602 and the expanded test data set 604 comprise data points, such as exemplary data point 606, relating to historical data from sampled medical patients.
  • the data set 600 may be used to train an SVM to learn whether a breast cancer patient will experience a recurrence of the disease.
  • each expanded data point includes twenty coordinates (or dimensions) 606al-3 through 606el-3, and an output classification 606f, which collectively represent medical data and categorization transformations thereof for each patient.
  • the first coordinate 606a represents "Age”
  • the second coordinate through the fourth coordinate 606a 1 - 606a3 are variables that combine to represent a category of age.
  • a range of ages may be categorized, for example, into “young" "middle-aged” and “old” categories respective to the range of ages present in the data.
  • a string of variables "0" (606al), “0” (606a2), “1” (606a3) may be used to indicate that a certain age value is categorized as “old.”
  • a string of variables "0” (606al), “ 1” (606a2), “0” (606a3) may be used to indicate that a certain age value is categorized as “middle-aged.”
  • a string of variables "1" (606al), “0” (606a2), “0” (606al) may be used to indicate that a certain age value is categorized as
  • FIG. 7 illustrates an . exemplary expanded test output 702 from an SVM trained with the expanded training data set 602 and tested with the expanded data set 604 shown in FIG. 6.
  • the expanded test output 702 has been post-processed to be comprehensible by a human or computer.
  • the expanded test output 702 shows that 24 total samples (data points) were examined by the SVM and that the SVM incorrectly identified four of eight positive samples (50%) and incorrectly identified four of sixteen negative samples (25%). Accordingly, by comparing this expanded test output 702 with the unexpanded test output 502 of FIG. 5, it may be seen that the expansion of the data points leads to improved results (i.e. a lower global minimum error), specifically a reduced instance of patients who would unnecessarily be subjected to follow-up cancer treatments.
  • FIG. 8 illustrates an exemplary input and output for a stand alone application of the optimal categorization method 300 described in FIG. 3.
  • the input data set 801 comprises a "Number of Positive Lymph Nodes" 802 and a corresponding "Recurrence Classification" 804.
  • the optimal categorization method 300 has been applied to the input data set 801 in order to locate the optimal cutoff point for determination of treatment for cancer recurrence, based solely upon the number of positive lymph nodes collected in a post-surgical tissue sample.
  • the well-known clinical standard is to prescribe treatment for any patient with at least three positive nodes.
  • the optimal categorization method 300 demonstrates that the optimal cutoff 806, based upon the input data 801, should be at the higher value of 5.5 lymph nodes, which corresponds to a clinical rule prescribing follow-up treatments in patients with at least six positive lymph nodes.
  • the prior art accepted clinical cutoff point resulted in 47% correctly classified recurrences and 71% correctly classified non-recurrences. Accordingly, 53% of the recurrences were incorrectly classified (further treatment was improperly not recommended) and 29% of the non-recurrences were incorrectly classified (further treatment was incorrectly recommended).
  • the cutoff point determined by the optimal categorization method 300 resulted in 33% correctly classified recurrences and 97% correctly classified non- recurrences. Accordingly, 67% of the recurrences were incorrectly classified (further treatment was improperly not recommended) and 3% of the non -recurrences were incorrectly classified (further treatment was incorrectly recommended).
  • the exemplary optimal categorization method 300 it may be feasible to attain a higher instance of correctly identifying those patients who can avoid the post-surgical cancer treatment regimes, using the exemplary optimal categorization method 300. Even though the cutoff point determined by the optimal categorization method 300 yielded a moderately higher percentage of incorrectly classified recurrences, it yielded a significantly lower percentage of incorrectly classified non-recurrences. Thus, considering the trade-off, and realizing that the goal of the optimization problem was the avoidance of unnecessary treatment, the results of the cutoff point determined by the optimal categorization method
  • FIG. 9 is a comparison of exemplary post-processed output from a first support vector machine comprising a linear kernel and a second support vector machine comprising a polynomial kernel.
  • FIG. 9 demonstrates that a variation in the selection of a kernel may affect the level of quality of the output of an SVM.
  • the post- processed output of a first SVM 902 comprising a linear dot product kernel indicates that for a given test set of twenty-four samples, six of eight positive samples were incorrectly identified and three of sixteen negative samples were incorrectly identified.
  • the post-processed output for a second SVM 904 comprising a polynomial kernel indicates that for the same test set only two of eight positive samples were incorrectly identified and four of sixteen negative samples were identified.
  • the polynomial kernel yielded significantly improved results pertaining to the identification of positive samples and yielded only slightly worse results pertaining to the identification of negative samples.
  • the global minimum error for the polynomial kernel is lower than the global minimum error for the linear kernel for this data set.
  • FIG. 10 and the following discussion are intended to provide a brief and general description of a suitable computing environment for implementing the present invention.
  • the computer 1000 includes a central processing unit 1022, a system memory 1020, and an Input/Output ("I/O") bus 1026.
  • a system bus 1021 couples the central processing unit 1022 to the system memory 1020.
  • a bus controller 1023 controls the flow of data on the I/O bus 1026 and between the central processing unit 1022 and a variety of internal and external I/O devices.
  • the I/O devices connected to the I/O bus 1026 may have direct access to the system memory 1020 using a Direct Memory Access (“DMA”) controller 1024.
  • DMA Direct Memory Access
  • the I/O devices are connected to the I/O bus 1026 via a set of device interfaces.
  • the device interfaces may include both hardware components and software components.
  • a hard disk drive 1030 and a floppy disk drive 1032 for reading or writing removable media 1050 may be connected to the I/O bus 1026 through disk drive controllers 1040.
  • An optical disk drive 1034 for reading or writing optical media 1052 may be connected to the I/O bus 1026 using a Small Computer System Interface ("SCSI") 1041.
  • SCSI Small Computer System Interface
  • an IDE (ATAPI) or EIDE interface may be associated with an optical drive such as a may be the case with a CD-ROM drive.
  • the drives and their associated computer-readable media provide nonvolatile storage for the computer 1000.
  • other types of computer-readable media may also be used, such as ZIP drives, or the like.
  • a display device 1053 such as a monitor, is connected to the I/O bus 1026 via another interface, such as a video adapter 1042.
  • a parallel interface 1043 connects synchronous peripheral devices, such as a laser printer 1056, to the I/O bus 1026.
  • a serial interface 1044 connects communication devices to the I/O bus 1026.
  • a user may enter commands and information into the computer 1000 via the serial interface 1044 or by using an input device, such as a keyboard 1038, a mouse 1036 or a modem 1057.
  • peripheral devices may also be connected to the computer 1000, such as audio input/output devices or image capture devices.
  • a number of program modules may be stored on the drives and in the system memory 1020.
  • the system memory 1020 can include both Random Access Memory (“RAM”) and Read Only Memory (“ROM”).
  • the program modules control how the computer 1000 functions and interacts with the user, with I/O devices or with other computers.
  • Program modules include routines, operating systems 1065, application programs, data structures, and other software or firmware components.
  • the present invention may comprise one or more pre-processing program modules 1075A, one or more post-processing program modules 1075B, and/or one or more optimal categorization program modules 1077 and one or more SVM program modules 1070 stored on the drives or in the system memory 1020 of the computer 1000.
  • pre-processing program modules 1075A, post-processing program modules 1075B, together with the SVM program modules 1070 may comprise computer-executable instructions for pre-processing data and post-processing output from a learning machine and implementing the learning algorithm according to the exemplary methods described with reference to FIGS. 1 and 2.
  • optimal categorization program modules 1077 may comprise computer-executable instructions for optimally categorizing a data set according to the exemplary methods described with reference to FIG. 3.
  • the computer 1000 may operate in a networked environment using logical connections to one or more remote computers, such as remote computer 1060.
  • the remote computer 1060 may be a server, a router, a peer device or other common network node, and typically includes many or all of the elements described in connection with the computer 1000.
  • program modules and data may be stored on the remote computer 1060.
  • the logical connections depicted in FIG. 10 include a local area network ("LAN”) 1054 and a wide area network (“WAN”) 1055.
  • a network interface 1045 such as an Ethernet adapter card, can be used to connect the computer 1000 to the remote computer 1060.
  • the computer 1000 may use a telecommunications device, such as a modem 1057, to establish a connection.
  • a telecommunications device such as a modem 1057
  • the network connections shown are illustrative and other devices of establishing a communications link between the computers may be used.
  • FIG. 11 is a functional block diagram illustrating an alternate exemplary operating environment for implementation of the present invention.
  • the present invention may be implemented in a specialized configuration of multiple computer systems.
  • An example of a specialized configuration of multiple computer systems is referred to herein as the BlOWulfTM Support Vector Processor (BSVP).
  • the BSVP combines the latest advances in parallel computing hardware technology with the latest mathematical advances in pattern recognition, regression estimation, and density estimation. While the combination of these technologies is a unique and novel implementation, the hardware configuration is based upon Beowulf supercomputer implementations pioneered by the NASA Goddard Space Flight Center.
  • the BSVP provides the massively parallel computational power necessary to expedite SVM training and evaluation on large-scale data sets.
  • the BSVP includes a dual parallel hardware architecture and custom parallelized software to enable efficient utilization of both multithreading and message passing to efficiently identify support vectors in practical applications. Optimization of both hardware and software enables the BSVP to significantly outperform typical SVM implementations.
  • the upgradability of the BSVP is ensured by its foundation in open source software and standardized interfacing technology. Future computing platforms and networking technology can be assimilated into the
  • the BSVP as they become cost effective with no effect on the software implementation.
  • the BSVP comprises a Beowulf class supercomputing cluster with twenty processing nodes 1104a-t and one host node 1112.
  • the processing nodes 1104a-j are interconnected via switch 1102a, while the processing nodes 1104k-t are interconnected via switch 1102b.
  • Host node 1112 is connected to either one of the network switches 1102a or 1102b (1102a shown) via an appropriate Ethernet cable
  • switch 1102a and switch 1102b are connected to each other via an appropriate Ethernet cable 1114 so that all twenty processing nodes 1104a-t and the host node 1112 are effectively in communication with each other.
  • Switches 1102a and 1102b preferably comprise Fast Ethernet interconnections.
  • the dual parallel architecture of the BSVP is accomplished through implementation of the Beowulf supercomputer's message passing multiple machine parallel configuration and utilizing a high performance dual processor SMP computer as the host node 1112.
  • the host node 1112 contains glueless multi -processor SMP technology and consists of a dual 450Mhz Pentium II Xeon based machine with 18GB of Ultra SCSI storage, 256MB memory, two lOOMbit/sec
  • the host node 1112 executes NIS, MPL and/or PVM under Linux to manage the activity of the BSVP.
  • the host node 1112 also provides the gateway between the BSVP and the outside world. As such, the internal network of the BSVP is isolated from outside interaction, which allows the entire cluster to appear to function as a single machine.
  • the twenty processing nodes 1104a-t are identically configured computers containing 150MHz Pentium processors, 32MB RAM, 850MB HDD, 1.44MB FDD, and a Fast Ethernet mblOOMb/s NIC.
  • the processing nodes 1104a-t are interconnected with each other and the host node through NFS connections over TCP/IP.
  • the processing nodes are configured to provide demonstration capabilities through an attached bank of monitors with each node's keyboard and mouse routed to a single keyboard device and a single mouse device through the KVM switches 1108a and 1108b.
  • the software implements full cycle support from raw data to implemented solution.
  • a database engine provides the storage and flexibility required for preprocessing raw data.
  • Custom developed routines automate the pre-processing of the data prior to SVM training.
  • Multiple transformations and data manipulations are performed within the database environment to generate candidate training data.
  • the peak theoretical processing capability of the BSVP is 3.90GFLOPS. Based upon the benchmarks performed by NASA Goddard Space Flight Center on their Beowulf class machines, the expected actual performance should be about 1.56GFLOPS.
  • the massive computing power of the BSVP renders it particularly useful for implementing multiple SVMs in parallel to solve real-life problems that involve a vast number of inputs.
  • Examples of the usefulness of SVMs in general and the BSVP in particular comprise: genetic research, in particular the Human Genome Project; evaluation of managed care efficiency; therapeutic decisions and follow up; appropriate therapeutic triage; pharmaceutical development techniques; discovery of molecular structures; prognostic evaluations; medical informatics; billing fraud detection; inventory control; stock evaluations and predictions; commodity evaluations and predictions; and insurance probability estimates.
  • FIG. 12 is a functional block diagram illustrating an exemplary network operating environment for implementation of a further alternate embodiment of the present invention.
  • a customer 1202 or other entity may transmit data via a distributed computer network, such as the Internet 1204, to a vendor 1212.
  • a distributed computer network such as the Internet 1204
  • the customer 1202 may transmit data from any type of computer or lab instrument that includes or is in communication with a communications device and a data storage device.
  • the data transmitted from the customer 1202 may be training data, test data and/or Uve data to be processed by a learning machine.
  • the data transmitted by the customer is received at the vendor's web server 1206, which may transmit the data to one or more learning machines via an internal network 1214a-b.
  • learning machines may comprise SVMs, BSVPs 1100, neural networks, other learning machines or combinations thereof.
  • the web server 1206 is isolated from the learning machine(s) by way of a firewall 1208 or other security system.
  • the vendor 1212 may also be in communication with one or more financial institutions 1210, via the Internet 1204 or any dedicated or on-demand communications link.
  • the web server 1206 or other communications device may handle communications with the one or more financial institutions.
  • the financial institution(s) may comprise banks, Internet banks, clearing houses, credit or debit card companies, or the like.
  • the vendor may offer learning machine processing services via a web-site hosted at the web-server 1206 or another server in communication with the web-server 1206.
  • a customer 1202 may transmit data to the web server 1206 to be processed by a learning machine.
  • the customer 1202 may also transmit identification information, such as a username, a password and/or a financial account identifier, to the web-server.
  • the web server 1206 may electronically withdraw a pre-determined amount of funds from a financial account maintained or authorized by the customer 1202 at a financial institution 1210.
  • the web server may transmit the customer's data to the BSVP 1100 or other learning machine.
  • the post-processed output is returned to the webserver 1206.
  • the output from a learning machine may be post- processed in order to generate a single-valued or multi-valued, computationally derived alpha-numerical classifier, for human or automated interpretation.
  • the web server 1206 may then ensure that payment from the customer has been secured before the post- processed output is transmitted back to the customer 1202 via the Internet 1204.
  • SVMs may be used to solve a wide variety of real-life problems.
  • SVMs may have applicability in analyzing accounting and inventory data, stock and commodity market data, insurance data, medical data, etc.
  • the above-described network environment has wide applicability across many industries and market segments.
  • a customer may be a retailer.
  • the retailer may supply inventory and audit data to the web server 1206 at predetermined times.
  • the inventory and audit data may be processed by the BSVP and/or one or more other learning machine in order to evaluate the inventory requirements of the retailer.
  • the customer may be a medical laboratory and may transmit live data collected from a patient to the web server 1206 while the patient is present in the medical laboratory.
  • the output generated by processing the medical data with the BSVP or other learning machine may be transmitted back to the medical laboratory and presented to the patient.
  • data input is a vector called a "pattern" of components called “features”.
  • features are gene expression coefficients and patterns correspond to patients.
  • a two-class classification problem is shown.
  • a training set of a number of patterns with known class labels was used.
  • the training patterns were used to build a decision function or a discriminant function that is a scalar function of an input pattern. New patterns are classified according to the sign of the decision function.
  • Decision functions that are simple weighted sums of the training patterns plus a bias are called linear discriminiant functions.
  • a data set is said to be “linearly separable” if a linear discriminant function can separate it without error.
  • a known problem in classification, and machine learning in general, is to find means to reduce the dimensionality of input space to overcome the risk of "overfitting".
  • Other methods of reduction include projecting on the first few principal directions of the data. With such method, new features are obtained that are linear combinations of the original features.
  • projection methods One disadvantage of projection methods is that none of the original input features can be discarded.
  • Preferred methods comprise pruning techniques that ehminate some of the original input features and retain a minimum subset of features that yield a better classification performance. For diagnostic tests, it is of practical importance to be able to select a small subset of genes for reasons such as cost effectiveness and so that the relevance of the genes selected can be verified more easily.
  • This method is impractical for large numbers of features, such as thousands of genes, because of the combinatorial explosion of the number of subsets.
  • Performing feature selection in large dimensional input spaces involves greedy algorithms.
  • feature ranking techniques are particularly preferred.
  • a fixed number of top ranked features may be selected for further analysis or to design a classifier.
  • a threshold can be set on the ranking criterion. Only the features whose criterion exceed the threshold are retained.
  • a preferred method is to use the ranking to define nested subsets of features and select an optimum subset of features with a model selection criterion by varying a single parameter: the number of features.
  • the present invention also comprises methods, systems and devices of multiple support vector machines for discovering knowledge from multiple data sets.
  • a plurality of support vector machines may be configured to hierarchically process multiple data sets in parallel or in sequence.
  • one or more first-level support vector machines may be trained and tested to process a first type of data and one or more first-level support vector machines may be trained and tested to process a second type of data. Additional types of data may be processed by other first-level support vector machines as well.
  • the output from some or all of the first-level support vector machines may be combined in a logical manner so as to produce an input data set for one or more second-level support vector machines.
  • output from a plurality of second-level support vector machines may be combined in a logical manner to produce input data for one or more third-level support vector machine.
  • the hierarchy of support vector machines may be expanded to any number of levels as may be appropriate.
  • Each support vector machine in the hierarchy or each hierarchical level of support vector machines may be configured with a distinct kernel.
  • support vector machines used to process a first type of data may be configured with a first type of kernel
  • support vector machines used to process a second type of data may be configured with a second type of kernel.
  • multiple support vector machines in the same or different hierarchical level may be configured to process the same type of data using distinct kernels.
  • a first-level support vector machine may be trained and tested to process mammography data pertaining to a sample of medical patients.
  • An additional first-level support vector machine may be trained and tested to process genomic data for the same or a different sample of medical patients.
  • the output from the two first-level support vector machines may be combined to form a new multi-dimensional data set relating to mammography and genomic data.
  • the new data set may then be processed by an appropriately trained and tested second- level support vector machine.
  • the resulting output from the second-level support vector machine may identify causal relationships between the mammography and genomic data points.
  • the contemplated hierarchy of support vector machines may have applications in any field or industry in which analysis of data by a learning machine is desired.
  • the hierarchical processing of multiple data sets using multiple support vector machines may be used as a method for pre-processing or post-processing data that is to be input to or output from still other support vector machines or learning machines.
  • pre-processing or post-processing of data according to the methods described below may be performed to the input data and/or output of the above-described hierarchical architecture of support vector machines.
  • FIG. 36 is presented by way of example only to illustrate a hierarchical system of support vector machines.
  • one or more first-level support vector machines 1302A1 and 1302A2 may be trained and tested to process a first type of input data 1304A, such as mamography data, pertaining to a sample of medical patients.
  • a first type of input data 1304A such as mamography data
  • One or more of these support vector machines may comprise a distinct kernel (shown as kernel 1 and kernel 2).
  • one or more additional first-level support vector machines 1302B1 and 1302B2 may be trained and tested to process a second type of data 1304B, such as genomic data, for the same or a different sample of medical patients.
  • the additional support vector machines may comprise a distinct kernel (shown as kernel 1 and kernel 3).
  • the output from each of the like first level support vector machines may be compared with each other (i.e., output Al 1306A compared with output A2 1306B; output Bl 1306C compared with output B2 1306D) in order to determine optimal outputs (1308A and 1308B). Then, the optimal outputs from the two types of first-level support vector machines 1308A and 1308B may be combined to form a new multi-dimensional input data set 1310, for example relating to mamography and genomic data. The new data set may then be processed by one or more appropriately trained and tested second-level support vector machines 1312A and 1312B.
  • the resulting outputs 1314A and 1314B from the second-level support vector machines 1312A and 1312B may be compared to determine an optimal output 1316.
  • the optimal output 1316 may identify causal relationships between the mamography and genomic data points.
  • the contemplated hierarchy of support vector machines may have applications in any field or industry in which analysis of data by a learning machine is desired.
  • the hierarchical processing of multiple data sets using multiple support vector machines may be used as a method for pre-processing or post-processing data that is to be input to or output from still other support vector machines or learning machines.
  • pre-processing or post-processing of data may be performed to the input data and/or output of the above-described hierarchical architecture of support vector machines.
  • the examples included herein show preferred methods for determining the genes that are most correlated to the presence of colon cancer or can be used to predict colon cancer occurance in an individual.
  • the present invention comprises these methods, and other methods, including other computational methods, usable in a learning machine for determining genes, proteins or other measurable criteria for the diagnosis or prognosis of changes in a biological system.
  • the preferred optimum number of genes is a range of approximately from 1 to 100, more preferably, the range is from 1 to 50, even more preferably the range is from 1 to 32, still more preferably the range is from 1 to 21 and most preferably, from 1 to 10.
  • the preferred optimum number of genes can be affected by the quality and quantity of the original data and thus can be determined for each application by those skilled in the art.
  • therapeutic agents can be administered to antagonize or agonize, enhance or inhibit activities, presence, or synthesis of the gene products.
  • Therapeutic agents include, but are not limited to, gene therapies such as sense or antisense polynucleotides, DNA or RNA analogs, pharmaceutical agents, plasmaphoresis, antiangiogenics, and derivatives, analogs and metabolic products of such agents.
  • Such agents are administered via parenteral or noninvasive routes.
  • Many active agents are administered through parenteral routes of administration, intravenous, intramuscular, subcutaneous, intraperitoneal, intraspinal, intrathecal, intracerebroventricular, intraarterial and other routes of injection.
  • Noninvasive routes for drug delivery include oral, nasal, pulmonary, rectal, buccal, vaginal, transdermal and occular routes.
  • Another embodiment of the present invention comprises use of testing remote from the site of determination of the patterns through means such as the internet or telephone lines. For example, a genomic test to identify the presence of genes known to be related to a specific medical condition is performed in a physician's office. Additionally, other information such as clinical data or proteomic determinations may also be made at the same time or a different time. The results of one, some or all of the tests are transmitted to a remote site that houses the SVMs. Such testing could be used for the diagnosis stages, for determining the prognosis of the disease, for determining the results of therapy and for prescriptive applications such as determining which therapy mitine is better for individual patients.
  • Errorless separation can be achieved with any number of genes, from one to many.
  • Preferred methods comprise use of larger numbers of genes.
  • Classical gene selection methods select the genes that individually classify the training data best. These methods include correlation methods and expression ratio methods. They ehminate genes that are useless for discrimination (noise), but do not yield compact gene sets because genes are redundant. Moreover, complementary genes that individually do not separate well the data are missed.
  • a simple feature (gene) ranking can be produced by evaluating how well an individual feature contributes to the separation (e.g. cancer vs. normal).
  • Various correlation coefficients are used as ranking criteria. The coefficient used is defined as:
  • a preferred method for the present invention comprises using the gene ranking
  • weights multiplying the inputs of a given classifier can be used as gene ranking coefficients.
  • the inputs that are weighted by the largest values have the most influence in the classification decision. Therefore, if the classifier performs weU, those inputs with largest weights correspond to the most informative genes.
  • Other methods comprise algorithms to train linear discriminant
  • a preferred method of the present invention is to use the weights of a classifier to produce a feature ranking with an SVM (Support Vector Machine).
  • SVM Small Vector Machine
  • the present invention contemplates methods of SVMs used for non-linear decision boundaries of
  • FIG. 13 graphically illustrates use of a linear discriminant classifier.
  • the x and y coordinates represent the expression coefficients of two genes.
  • a linear discriminant classifier makes its decision according to the sign of a weighted sum of the x and y
  • SVMs are maximum margin classifiers in their input components. See Fig. 13-a and 13-b. The decision boundary (a straight line in the case of a two-dimensional separation) is positioned to leave the largest possible margin on either side. A particularity of SVMs
  • FIG. 13-c and 13-d show separation of the training examples with an SVM. The training examples are separated without error. The margin on either side of the decision boundary is maximized. 13b shows separation of the training and test examples with the same SVM. Only one example is misclassified. 13c shows separation of the training examples with the baseline method of Golub, 1999. The decision boundary is perpendicular to the direction defined by the class centroids. 13d shows separation of the training and test examples with the baseline method. Three examples are misclassified.
  • ⁇ ot, ⁇ C and ⁇ , ot, y, 0
  • x that are vectors of features (genes)
  • x r x denotes the scalar product
  • y encodes the class label as a binary value +1 or -1
  • ⁇ and C are positive constants (soft margin parameters).
  • the soft margin parameters ensure convergence even when the problem is non-linearly separable or poorly conditioned. In such cases, some of the support vectors may not lie on the margin.
  • the weight vector w is a linear combination of training patterns. Most weights oc, are zero.
  • the training patterns with non zero weights are support vectors. Those with weight satisfying the strict inequality 0 ⁇ c., ⁇ C are marginal support vectors.
  • the bias value i is an average over marginal support vectors.
  • the original data for training and testing the learning machine of the present invention for this application regarding colon cancer was derived from the data presented in Alon et al., 1999.
  • Gene expression information was extracted from microarray data resulting, after pre-processing, in a table of 62 tissues x 2000 genes.
  • the 62 tissues include 22 normal tissues and 40 colon cancer tissues.
  • the matrix contains the expression of the 2000 genes with highest minimal intensity across the 62 tissues.
  • One problem in the colon cancer data set was that tumor samples and normal samples differed in cell composition. Tumor samples were normally rich in epithelial cells wherein normal samples were a mixture of cell types, including a large fraction of smooth muscle cells. Though the samples could be easily separated on the basis of cell composition, this separation was not very informative for tracking cancer-related genes.
  • Alon et al. provides an analysis of the data based on top down clustering, a method of unsupervised learning and also clusters genes by showing that some genes correlate with a cancer vs normal separation scheme but do not suggest a specific method of gene selection. They show that some genes are correlated with the cancer vs. normal separation but do not suggest a specific method of gene selection.
  • the gene selection method of this embodiment of present invention comprises a reference gene selection method like that of Example 2 and like that used in Golub et al, Science, 1999.
  • Golub the authors use several metrics of classifier quality, including error rate, rejection rate at fixed threshold, and classification confidence.
  • Each value is computed both on the independent test set and using the leave-one-out method on the training set.
  • the leave-one-out method consists of removing one example from the training set, constructing the decision function on the basis only of the remaining training data and then testing on the removed example. In this method, one tests all examples of the training data and measures the fraction of error over the total number of training examples.
  • the methods of using the learning machine comprise modifications of the above metrics.
  • the classification decision was carried out according to the sign of the SVM output. The magnitude of the output is indicative of classification confidence.
  • Error (B 1+B2) number of errors ("bad") at zero rejection.
  • Reject (R1+R2) minimum number of rejected samples to obtain zero error.
  • Extremal margin (E/D) difference between the smallest output of the positive class samples and the largest output of the negative class samples (rescaled by the largest difference between outputs).
  • Median margin (M/D) difference between the median output of the positive class samples and the median output of the negative class samples (rescaled by the largest difference between outputs).
  • the error rate is the fraction of examples that are misclassified (corresponding to a diagnostic error). It is complemented by the success rate.
  • the rejection rate is the fraction of examples that are rejected (on which no decision is made because of low confidence). It is complemented by the acceptance rate.
  • Extremal and median margins are measurements of classification confidence. This method of computing the margin with the leave-one-out method or on the test set differed from the margin computed on training examples sometimes used in model selection criteria.
  • a method for predicting the optimum subset of genes comprised defining a criterion of optimality that uses information derived from training examples only. This was checked by determining whether the predicted gene subset performed best on the test set.
  • a criterion that is often used in similar "model selection" problems is the leave- one-out success rate V suc .
  • V suc the success rate
  • V acc the acceptance rate
  • V ext the extremal margin
  • V med the median margin
  • RFE SVM Recursive Feature Elimination
  • the leave-one-out method with the classifier quality criterion was used to estimate the optimum number of genes.
  • Example 2 also illustrates use of the leave-one- out method.
  • the leave-one-out method comprises taking out one example of the training set. Training is performed on the remaining examples. The left out example is used to test. The procedure is iterated over all the examples. Every criteria is computed as an average over all examples.
  • the overall classifier quality criterion is the sum of 4 values: the leave-one-out success rate (at zero rejections), the leave-one-out acceptance rate (at zero error), the leave-one-out extremal margin, and the leave-one-out median margin.
  • the classifier is a linear classifier with hard margin.
  • Figure 15 shows the distributions of gene expression values across tissue samples for two random genes (cumulative number of samples of a given expression value) which is compared with a uniform distribution. Each line represents a gene. 15A and B show the raw data; 15 C and D are the same data after taking the log. By taking the log of the gene expression values the same curves result and the distribution is more uniform. This may be due to the fact that gene expression coefficients are often obtained by computing the ratio of two values. For instance, in a competitive hybridization scheme,
  • DNA from two samples that are labeled differently are hybridized onto the array.
  • the first initial preprocessing step that is taken is to take the ratio a/b of these two values. Though this initial preprocessing step is adequate, it may not be optimal when the two values are small.
  • Other initial preprocessing steps include (a-b)/(a+b) and (log a - log b)/(log a + log b).
  • Figure 16 shows the distribution of gene expression values across genes for aU tissue samples.
  • 16A shows the raw data and 16B shows the inv erf.
  • the shape is roughly that of an erf function, indicating that the density follows approximately the Normal law. Indeed, passing the data through the inverse erf function yields almost straight parallel lines. Thus, it is reasonable to normalize the data by subtracting the mean.
  • This preprocessing step is also suggested by Alon et al. This preprocessing step is supported by the fact that there are variations in experimental conditions from microarray to microarray. Although standard deviation seems to remain fairly constant, the other preprocessing step selected was to divide the gene expression values by the standard deviation to obtain centered data of standardized variance.
  • Figure 17 shows the results of these preprocessing steps.
  • Figure 17 shows the data matrices representing gene expression values from microarray data for colon cancer wherein the lines represent 62 tssues and the columns represent the 2000 genes.
  • an additional preprocessing step was added by passing the data through a squashing function to diminish the importance of the outliers.
  • Example 2 The data was preprocessed as described above and summarized in Figure 17 to produce new and improved results. In this method, there are modifications from those used in Example 2.
  • the code was optimized such that RFE can be run by eliminating one gene at a time. In Example 2, chunks of genes at a time were eliminated. The chunk size was divided by 2 at every iteration.
  • This processing modification of this embodiment provides a finer ranking that allows for various analyses but does not significantly affect classification accuracy. It runs in about 10-15 minutes, for example, on a Pentium III 333, 256 MB RAM. .
  • a reduced capacity SVM was used by projecting first the data on the first principal component.
  • Figure 18 shows the results of RFE after preprocessing.
  • Reduced capacity SVM used in Figure 14 is replaced by plain SVM. Although a log scale is still used for gene number, RFE was run by eliminating one gene at a time. The best test performance is 90% classification accuracy (8 genes). The optimum number of genes predicted from the classifier quality based on training data information only is 16. This corresponds to 87% classification accuracy on the test set. The same test performance is also achieved with only 2 genes as follows:
  • J02854 Myosin regulatory light chain 2, smooth muscle isoform human); contains element TAR 1 repetitive element.
  • the top gene found is a smooth muscle gene, which is a gene characteristic of tissue composition and is probably not related to cancer.
  • Golub's gene selection method is a ranking method where genes are ordered according to the correlation between vectors of gene expression values for all training data samples and the vector of target values (+1 for normal sample and -1 for cancer sample). Golub et al select m/2 top ranked and m/2 bottom ranked genes to obtain one half of genes highly correlated with the separation and one half anti-correlated. Golub et al use a linear classifier. To classify an unknown sample, each gene "votes" for cancer or normal according to its correlation coefficient with the target separation vector. The top gene selected by Golub's method was J02854 (smooth muscle related).
  • Table 1 Error rates comparisons between the methods of this embodiment of the present invention and Golub's method. The list of errors is shown between brackets. The numbers indicate the patients. The sign indicates cancer (negative) or normal (positive). For this embodiment of the present invention, the best performance was at 8 genes and the optimum predicted at 16 genes. For Golub, the best performance was at 16 genes and the optimum predicted at 4 genes. Note that there was only one error difference between the best performance and the optimum predicted in either case. Combining clustering and gene selection
  • RFE SVM recursive feature elimination
  • Correlation methods such as Golub's method provide a ranked list of genes.
  • the rank order characterizes how correlated the gene is with the separation.
  • a gene highly ranked taken alone provides a better separation than a lower ranked gene. It is therefore possible to set a threshold (e.g. keep only the top ranked genes) that separates "highly informative genes” from “less informative genes”.
  • the methods of the present invention such as SVM RFE provide subsets of genes that are both smaller and more discriminant.
  • the SVM gene selection method using RFE also provides a ranked list of genes. With this list, nested subsets of genes of increasing sizes can be defined. However, the fact that one gene has a higher rank than another gene does not mean that that factor alone characterizes the better separation. In fact, genes that are eliminated very early on may be very informative but redundant with others that were kept.
  • the absolute values of correlation coefficients are larger between the 32 best genes and the other genes that have highest rank.
  • 20B shows the SVM method.
  • the 32 best genes as a whole provide a good separation but individually may not be very correlated with the target separation.
  • Gene ranking allows for a building nested subsets of genes that provide good separations. It is not informative for how good an individual gene is. Genes of any rank may be correlated with the 32 best genes. They may been ruled out at some point because the their redundancy with some of the remaining genes, not because they did not carry information relative to the target separation.
  • the gene ranking alone is insufficient to characterize which genes are informative and which ones are not, and also to determine which genes are complementary and which ones are redundant. Unsupervised clustering
  • the data was preprocessed with an unsupervised clustering method. Genes were grouped according to resemblance (with a given metric). Cluster centers are then used instead of genes themselves and processed by SVM RFE. The result was nested subsets of cluster centers. An optimum subset size can be chosen with the same cross-validation method used before. The cluster centers can then be replaced either element of the cluster.
  • Figure 21 shows the performance curves.
  • Figure 21 shows the results of RFE when training on 100 dense QT_clust clusters.
  • Horizontal axis log2 (number of gene cluster centers).
  • Figure 22 shows the top 8 QT_clust clusters chosen by SVM RFE.
  • the gene expression for the 32 tissues of the training set (columns) for 8 clusters (lines) are represented. Positive gene expressions are red and negative gene expressions are blue. Small values have lighter color.
  • 22A shows cluster centers; 22B shows cluster elements.
  • the cluster centers can be substituted by any of their members. This factor may be important in the design of some medical diagnosis tests. For example, the administration of some proteins may be easier than that of others. Having a choice of alternative genes introduces flexibility in the treatment and administration choices.
  • Hierarchical clustering instead of QT_clust clustering was used to produce lots of small clusters containing 2 elements on average. Because of the smaller cluster cardinality, there were fewer gene alternatives from which to choose. In this instance, hierarchical clustering did not yield as good a result as using QT_clust clustering.
  • the present invention contemplates use of any of the known methods for clustering, including but not limited to hierarchical clustering, QT_clust clustering and SVM clustering.
  • the choice of which clustering method to employ in the invention is affected by the initial data and the outcome desired, and can be determined by those skilled in the art.
  • Each dot represent the gene expression value of average patients obtained by principal component analysis.
  • the colored dots are the genes selected by SVM RFE using QT_clust clusters. Each cluster is assigned a randomly selected color. The dot size is proportional to the cluster rank.
  • all normal tissue was replaced by a single average normal tissue (first principal component called "principal normal tissue"). The same was done for the cancer tissues.
  • Each point represents a gene expression in the principal cancer tissue/principal normal tissue two-dimensional space.
  • Another method used with the present invention was to use clustering as a postprocessing step of SVM RFE.
  • Each gene selected by running regular SVM RFE on the original set of gene expression coefficients was used as a cluster center. For example, the results described in Figure 18 were used.
  • the correlation coefficient was computed with all remaining genes. The parameters were that the genes clustered to gene i were the genes that met the following two conditions: must have higher correlation coefficient with gene i than with other genes in the selected subset of eight genes, and must have correlation coefficient exceeding a threshold ⁇ .
  • FIG. 24 shows the gene expression for the 32 tissues of the training set (columns) for 8 clusters (lines. Positive gene expressions are red and negative gene expressions are blue. Small values have lighter color.
  • 24A shows the top 8 genes obtained by regular SVM RFE are used as cluster centers.
  • 24B shows all the elements of the clusters. Cluster elements may be highly correlated or anti-correlated to the cluster center.
  • the supervised clustering method does not give better control over the number of examples per cluster. Therefore, this method is not as good as unsupervised clustering if the goal is to be able to select from a variety of genes in each cluster.
  • supervised clustering may show specific clusters that have relevance for the specific knowledge being determined.
  • a very large cluster of genes was found that contained several muscle genes that may be related to tissue composition and may not be relevant to the cancer vs. normal separation. Thus, those genes are good candidates for elimination from consideration as having little bearing on the diagnosis or prognosis for colon cancer.
  • the following method was directed to eliminating the identified tissue composition related genes automaticaUy. Those genes complicate the analysis of the results because it was not possible to differentiate them from genes that are informative for the cancer vs. normal separation.
  • the results with the unsupervised learning preprocessing showed that the top ranked genes did not contain the key words "smooth muscle” that were used to detect potential tissue composition related genes. A cardiac muscle gene was still selected under this method.
  • the gene selection process was performed on the entire data set.
  • the learning machine such as the SVM used here, factored out tissue composition related genes.
  • SVM property of focusing on borderline cases may take advantage of a few examples of cancer tissue rich in muscle cells and of normal tissues rich in epithelial cells (the inverse of the average trend).
  • the resulting top ranking genes were free of muscle related genes, including the genes that were clustered with supervised clustering. In contrast, Golub's method obtains 3 smooth muscle related genes in the 7 top ranking gene cluster alone. Further, the top ranking genes found by SVM RFE were all characterizing the separation, cancer vs. normal (Table 4).
  • the present invention is not only making a quantitative difference on this data set with better classification accuracy and smaller gene subset, but is also making a qualitative difference: the gene set is free of tissue composition related genes.
  • Table 4 The 7 top ranked genes discovered by the methods of the present invention, in order of increasing importance.
  • Rk rank.
  • Sgn sign of correlation with the target separation, - for over-expressed in most cancer tissues; + for over-expressed in most normal tissues;
  • GAN Gene Accession Number; The possible function is derived from a keyword search involving "colon cancer” or "cancer” and some words in the gene description.
  • Figure 25 shows the results of the methods of the present invention using SVM RFE after training on the whole data set.
  • the optimum number of genes predicted by our leave-one-out criterion is 32 genes in Figure 25.
  • a finer plot in the small number of genes area reveals an optimum at 21 genes.
  • Figure 27 shows the weighting coefficients of the support vectors (the "alpha' s") in the last 100 iterations of SVM RFE.
  • the alpha vectors have been normalized. It is interesting to see that the alphas do not vary much until the very last iterations.
  • the number of support vectors goes through a minimum at 7 genes for 7 support vectors.
  • the muscle index is a quantity computed by Alon et al on all samples that reflects the muscle cell contents of a sample. Most normal samples have a higher muscle index than tumor samples. However, the support vectors do not show any such trend.
  • Table 5 Muscle index of the support vectors of the SVM trained on the top 7 genes selected by SVM RFE. Samples with a negative sign are tumor tissues. Samples with positive signs are normal tissues. Samples were ranked in ordered of increasing muscle index. In most samples in the data set, normal tissues have higher muscle index than tumor tissues because tumor tissues are richer in epithelial (skin) cells. This was not the case for support vectors which show a mix of all possibilities.
  • Figure 28 shows the top ranked genes discovered by SVM RFE in order of increasing importance from left to right.
  • the gene expression of all 62 tissues (columns) for the 7 clusters (lines) are represented.
  • the top 22 tissues are normal, the 40 last ones are cancerous. Positive gene expressions are red and negative gene expressions are blue. Small values have lighter color.
  • 28A shows cluster centers.
  • 28B shows output of the SVM (weighted sum of the genes of A). The separation is errorless.
  • the genes of Figure 28 do not look as orderly as those of Figure 29 because they are individually less correlated with the target separation, although together they carry more information.
  • Figure 29 shows the 7 top ranked genes discovered by Golub's methods in order of increasing improtance from left to right.
  • the gene expression of all 62 tissues (columns) for the 7 clusters (lines) are represented.
  • the top 22 tissues are normal, the 40 last ones are cancerous.
  • Positive gene expressions are red and negative gene expressions are blue. Small values have lighter color.
  • 29A shows cluster centers.
  • 29B shows output of the Golub classifier (weighted sum of the genes of A). The separation is not errorless.
  • SVM RFE differed from Golub's method in two respects: the mutual information between features was used by SVMs while Golub's method makes implicit independence assumptions; and the decision function was based only on support vectors that are "borderline” cases as opposed to being based on aU examples in an attempt to characterize the "typical” cases.
  • the use of support vectors is critical in factoring out irrelevant tissue composition related genes.
  • SVM RFE was compared with RFE methods using other linear discriminant functions that do not make independence assumptions but attempt to characterize the "typical” cases. Two discriminant functions were chosen:
  • LDA Linear Discriminant Analysis
  • MSE Mean-Squared-Error
  • SVM RFE gives the best results down to 4 genes. An examination of the genes selected reveals that SVM eliminates genes that are tissue composition-related and keeps only genes that are relevant to the cancer vs. normal separation. Conversely, other methods keep smooth muscle genes in their top ranked genes which helps in separating most samples but is not relevant to the cancer vs. normal discrimination.
  • Figure 35 shows the selection of an optimum number of genes for colon cancer data.
  • the number of genes selected by recursive gene elimination with SVMs was varied.
  • the lines of the graph are as follows: Circle: error rate on the test set. Square: scaled quality criterion (Q/4). Crosses: scaled criterion of optimality (C/4). Diamond curve: result of locally smoothing the C/4.
  • the model selection criterion was established using leukemia data, its predictive power was correlated by using it on colon cancer data, without making any adjustment. The criterion also predicted the optimum accurately. The performance was not as accurate on that first trial because the same preprocessing as for the leukemia data of
  • Example 2 was used. The results were improved substantially by adding several preprocessing steps and reached a success rate of 90% accuracy. These preprocessing steps included taking the logarithm of aU values, normalizing sample vectors, normalizing feature vectors, and passing the result through a squashing function to diminish the importance of outliers. Normalization comprised subtracting the mean over all training values and dividing by the corresponding standard deviation.
  • the model selection criterion was used in a variety of other experiments using SVMs and other algorithms.
  • the optimum number of genes was always predicted accurately, within a factor of two of the number of genes. Results correlated with the biology literature
  • SVM RFE eliminated from its top ranked genes the smooth muscle genes that are likely to be tissue composition related.
  • the cancer related genes were limited to seven for convenience reasons. Additionally, the number seven corresponds to the minimum number of support vectors, a criterion also sometimes used for "model selection".
  • gene H64807 (Placental folate transporter) was identified as one of the most discriminant genes in the colon cancer vs. normal separation shows the use of the methods of the present invention for identifying genes involved in biological changes.
  • human chitotriosidase one needs to proceed by analogy with another homologous protein of the same family whose role in another cancer is under study: another chitinase (BRP39) was found to play a role in breast cancer. Cancer cells overproduce this chitinase to survive apoptosis (Aronson, 1999). Important increased chitotriosidase activity has been noticed in clinical studies of Gauchers disease patients, an apparently unrelated condition.
  • the chitotriosidase enzyme can be very sensitively measured.
  • the plasma or serum prepared from less than a droplet of blood is highly sufficient for the chitotriosidase measurement (Aerts, 1996). This opens the door to a possible new diagnosis test for colon cancer as well.
  • the 60S ribosomal protein L24 (Arabidopsis thaliana) is a non-human protein that is homologous a human protein located on chromosome 6. Like other ribosomal proteins, it may play a role in controlling cell growth and proliferation through the selective translation of particular classes of mRNA.
  • Trypanosoma is a parasitic protozoa indigenous to Africa and South America and patients infected by Trypanosoma (a colon parasite) develop resistance against colon cancer (Oliveira, 1999). Trypanosomiasis is an ancient disease of humans and animals and is still endemic in Africa and South America.
  • the data set which consisted of a matrix of gene expression vectors obtained from DNA microarrays, was obtained from cancer patients with two different types of leukemia.
  • the data set was easy to separate. After preprocessing, it was possible to find a weighted sum of a set of only a few genes that separated without error the entire data set, thus the data set was linearly separable. Although the separation of the data was easy, the problems present several features of difficulty, including small sample sizes and data differently distributed between training and test set.
  • the problem with the leukemia data was the distinction between two variants of leukemia (ALL and AML).
  • the data is split into two subsets: A training set, used to select genes and adjust the weights of the classifiers, and an independent test set used to estimate the performance of the system obtained.
  • Golub's training set consisted of 38 samples (27 ALL and 11
  • AML AML from bone marrow specimens.
  • Their test set has 34 samples (20 ALL and 14 AML), prepared under different experimental conditions and including 24 bone marrow and 10 blood sample specimens. All samples have 7129 attributes (or features) corresponding to some normalized gene expression value extracted from the micro-array image.
  • All samples have 7129 attributes (or features) corresponding to some normalized gene expression value extracted from the micro-array image.
  • the exact same experimental conditions were retained for ease of comparison with their method.
  • some of the large deviations between leave-one-out error and test error could not be explained by the small sample size alone.
  • the data analysis revealed that there are significant differences between the distribution of the training set and the test set. Various hypotheses were tested and it was found that the differences can be traced to differences in data source. In aU the experiments, the performance on test data from the various sources was followed separately. The results obtained were the same, regardless of the source.
  • the number of rejected examples Rl and R2 are the ordinates of - ⁇ R and ⁇ R in the triangle and circle curves respectively.
  • the decision function value of the rejected examples is smaller than ⁇ R in absolute value, which corresponds to examples of low classification confidence.
  • the threshold ⁇ R is set such that all the remaining "accepted" examples are well classified.
  • the extremal margin E is the difference between the smallest decision function value of class 2 examples and the largest decision function value of class 1 examples. On the example of the figure, E is negative. If the number of classification error is zero, E is positive.
  • the median margin M is the difference between the median decision function value of the class 1 density and the median of the class 2 density.
  • Table 8 Results of training classifiers on all genes (Leukemia data). A set of 50 genes corresponding to the largest weights of the SVM trained on all genes was selected. A new SVM was trained on these 50 genes . We compared the results with the baseline system trained with the original set of 50 features reported the Golub et al paper (Table 9).
  • a set of 50 genes was then selected.
  • the 50 genes corresponded to the largest weights of the SVM trained on all genes.
  • a new SVM was trained on these 50 genes.
  • Table 11 Detailed results of training on 50 genes (Leukemia data) The error id numbers are in brackets.
  • the classifiers trained on 50 genes are better than those trained on all genes with high confidence (based on the error rate 97.7% confidence for Golub and 98.7% for SVM). Based on the error rate alone, the SVM classifier is not significantly better than the Golub classifier (50% confidence on aU genes and 84.1% confidence on 50 genes). But, based on the rejections, the SVM classifier is significantly better than the Golub classifier (99.9% confidence on all genes and 98.7% confidence on 50 genes).
  • RFE Recursive Feature Elimination
  • a new classifier is trained with the remaining features.
  • the feature corresponding to the smallest weight in the new classifier is eliminated.
  • the order of elimination yields a particular ranking.
  • the last feature to be eliminated is ranked first. Chunks of genes were eliminated at a time. At the first iteration, the number of genes which is the closest power of 2 were reached. At subsequent iterations, half of the remaining genes were eliminated. Nested subsets of genes of increasing informative density were obtained.
  • the coefficients have been rescaled such that the average value of each indicator has zero mean an variance 1 across all four plots. For each classifier, the larger the colored area, the better the classifier. The figure shows that there is no significant difference between classifier performance on this data set, but there is a significant difference between the gene selections.
  • Table 12 Best classifier on test data (Leukemia data). The performance of the classifiers performing best on test data are reported. For each combination of SVM or Baseline genes and SVM or Baseline classifier, the corresponding number of genes, the number of errors and the number of rejections are shown in the table. The patient id numbers are shown in bracket.
  • Table 13 Fraction of common genes between the sets selected with the baseline method and SVM recursive gene elimination (Leukemia data). The fraction of common genes decreases approximately exponentially as a function of the number of genes (linearly in a log scale). Only 19% of the genes were common at the optimum SVM gene set number 16.
  • Figure 33 shows the best set of 16 genes for the leukemia data. In matrices (a) and (c), the columns represent different genes and the lines different patients from the training set. The 27 top lines are ALL patients and the 11 bottom lines are AML patients. The gray shading indicates gene expression: the lighter the stronger. 33A shows SVM best 16 genes. Genes are ranked from left to right, the best one at the extreme left. All the genes selected are more AML correlated.
  • 33B shows the weighted sum of the 16 SVM genes used to make the classification decision. A very clear ALL/AML separation is shown.
  • 33C shows baseline method 16 genes. The method imposes that half of the genes are AML correlated and half are ALL correlated. The best genes are in the middle.
  • 33D shows the weighted sum of the 16 baseline genes used to make the classification decision. The separation is still good, but not as good as the SVM separation. s
  • Figure 33A and 33C show the expression values for the patients in the training set of the 16 gene subsets. At first sight, the genes selected by the baseline method looked a lot more orderly. This was because they were strongly correlated with either
  • Table 15 Top ranked 16 baseline genes (Leukemia data).
  • Table 16 SVM classifier trained on SVM genes obtained with the RFE method (Leukemia data).
  • the criterion of classifier selection C was the classifier quality Q minus the error bar ⁇ . These quantities were computed based on training data only.
  • the success rate (at zero rejection), the acceptance rate (at zero error), the extreme margin and the median margin were reported for the leave-one-out method on the 38 sample training set (V results) and the 34 sample test set (T results). Where the number of genes was 16 was the best classifier predicted by the locally smoothed C criterion calculated using training data only.
  • Table 17 Best classifier selected with criterion C (Leukemia data). The performance of the classifiers corresponding to the optimum of criterion C, computed solely on the basis of training examples, were reported. For each combination of SVM or Baseline genes and SVM or Baseline classifier, the corresponding number of genes, the number of errors and the number of rejections are shown in the table. The patient id numbers are shown in bracket.
  • the leukemia data was treated by running the gene selection method on the entire data set of 72 samples.
  • the four top ranked genes are shown in Table 18.
  • Table 18 SVM RFE top ranked genes (Leukemia data). The entire data set of 72 samples was used to select genes with SVM RFE. Genes were ranked in order of increasing importance. The first ranked gene is the last gene left after all other genes have been eliminated. Expression: ALL>AML indicates that the gene expression level is higher in most ALL samples; AML>ALL indicates that the gene expression level is higher in most AML samples; GAN: Gene Accession Number. All the genes in this list have some plausible relevance to the AML vs. ALL separation.
  • the number of four genes corresponds the minimum number of support vectors (5 in this case). All four genes have some relevance to leukemia cancer and can be used for discriminating between AML and ALL variants.
  • the fastest methods of feature selection were correlation methods: for the data sets under study, several thousands of genes can be ranked in about one second by the baseline method of Golub with a Pentium processor.
  • the second fastest methods use the weights of a classifier trained only once with all the features as ranking criterion.
  • / Training algorithms such SVMs or Pseudo-inverse/MSE require first the computation of the (n,n) matrix K of all the scalar products between the n training patterns. The computation of K increases linearly with the number of features (gene) and quadratically with the number of training patterns. After that, the training time is of the order of the time to invert matrix K.
  • Recursive Feature Elimination requires training multiple classifiers on subsets of feature of decreasing size.
  • the training time scales linearly with number classifiers to be trained. Part of the calculations can be reused.
  • Matrix K does not need to be re-computed entirely.
  • the partial scalar products of the eliminated features can be subtracted.
  • the coefficients ⁇ can be initialized to their previous value.
  • the Matlab implementation of an SVM RFE of the present invention on a Pentium processor returns a gene ranking in about 15 minutes for the entire colon dataset (2000 genes, 62 patients) and 3 hours for the leukemia dataset (7129 genes, 72 patients). Given that the data collection and preparation may take several months or years, it is quite acceptable that the data analysis take a few hours.
  • SVM RFE improves feature selection based on feature ranking by eliminating the independence assumptions of correlation methods. It generates nested subsets of features. This means that the selected subset of d features is included in the subset of d+1 features. Feature ranking methods may miss a singleton that provides the best possible separation. These is no guarantee that the best feature pair will incorporate that singleton.
  • Combinatorial search is a computationally intensive alternative to feature ranking. To seek an optimum subset of d features or less, all combinations of d features or less are tried. The combination which yield the best classification performance is selected.
  • One embodiment of the present invention comprises using combinatorial methods.
  • a combinatorial search was used to refine the optimum feature set, starting with a subset of genes selected with SVM RFE.
  • the leukemia data was used in its training/test data split version.
  • the model selection criterion of the equation C Q-2 ⁇ (d). computed with the training dataset only, attempts were made to predict which combination would perform best on test data.,
  • the triplet of genes which ranked first provided 100% classification accuracy on both the training set and the test set.
  • Other embodiments of the present invention comprise use of non-linear classifiers.
  • SVM RFE also include use in problems of regression such as medical prognosis and for problems of density estimation or estimation of the support of a density.
  • RFE ranking can be thought of as producing nested subsets of features of increasing size that are optimal in some sense. Individually, a feature that is ranked better than another one may not separate the data better. In fact, there are features with any rank that are highly correlated with the first ranked feature.
  • One way of adding a correlation dimension to the simple linear structure provided by SVM RFE is to cluster genes according to a given correlation coefficient. Unsupervised clustering in pre-processing for SVM RFE was shown in the present application. The cluster centers were then used as features to be ranked. Supervised clustering was also used as a post-processing for SVM RFE. Top ranking features were also used as cluster centers. The remaining rejected features were clustered to those centers.
  • SVMs lend themselves particularly well to the analysis of broad patterns of gene expression from DNA microarray data. They can easily deal with a large number of features, such as thousands of genes, and a small number of training patterns, such as a small number of patients. Baseline methods were outperformed in only two days work by SVMs.
  • the two cancer databases showed that not taking into account the mutual information between genes in the process of selecting subsets of genes impairs classification performance. Significant improvements over the baseline methods that make imphcit independence assumptions were obtained. The top ranked genes found via SVM were all relevant to cancer. In contrast, other methods selected genes that were correlated with the separation at hand but were nor relevant to cancer diagnosis.
  • Angiostatin binds ATP synthase on the surface of human endothelial cells, PNAS, Vol. 96, Issue 6, 2811-2816, March 16, 1999, Cell Biology.
  • MSF MLF septin-like fusion
  • a fusion partner gene of MLL in a therapy-related acute myeloid leukemia with a t(ll;17)(q23;q25).

Abstract

L'invention concerne des procédés, des systèmes et des dispositifs qui mettent en oeuvre des appareils vectoriels de soutien servant à l'identification de modèles, lesquels sont importants pour poser un diagnotic ou un pronostic ou effectuer traitement médical. De tels modèles peuvent être trouvés dans plusieurs ensembles de données différents. L'invention concerne en outre des méthodes et des compositions servant à traiter ou à diagnostiquer des états pathologiques.
EP00973988A 1999-10-27 2000-10-27 Procedes et dispositifs pouvant identifier des modeles dans des systemes biologiques, et procedes d'utilisation Withdrawn EP1236173A2 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
EP10185728A EP2357582A1 (fr) 1999-10-27 2000-10-27 Procédés et dispositifs pour identifier les motifs de systèmes biologiques et leurs procédés d'utilisation

Applications Claiming Priority (15)

Application Number Priority Date Filing Date Title
US16180699P 1999-10-27 1999-10-27
US161806P 1999-10-27
US16870399P 1999-12-02 1999-12-02
US168703P 1999-12-02
US18459600P 2000-02-24 2000-02-24
US184596P 2000-02-24
US19121900P 2000-03-22 2000-03-22
US191219P 2000-03-22
US09/568,301 US6427141B1 (en) 1998-05-01 2000-05-09 Enhancing knowledge discovery using multiple support vector machines
US568301 2000-05-09
US09/578,011 US6658395B1 (en) 1998-05-01 2000-05-24 Enhancing knowledge discovery from multiple data sets using multiple support vector machines
US578011 2000-05-24
US20702600P 2000-05-25 2000-05-25
US207026P 2000-05-25
PCT/US2000/029770 WO2001031580A2 (fr) 1999-10-27 2000-10-27 Procedes et dispositifs pouvant identifier des modeles dans des systemes biologiques, et procedes d'utilisation

Publications (1)

Publication Number Publication Date
EP1236173A2 true EP1236173A2 (fr) 2002-09-04

Family

ID=27569077

Family Applications (2)

Application Number Title Priority Date Filing Date
EP00973988A Withdrawn EP1236173A2 (fr) 1999-10-27 2000-10-27 Procedes et dispositifs pouvant identifier des modeles dans des systemes biologiques, et procedes d'utilisation
EP10185728A Withdrawn EP2357582A1 (fr) 1999-10-27 2000-10-27 Procédés et dispositifs pour identifier les motifs de systèmes biologiques et leurs procédés d'utilisation

Family Applications After (1)

Application Number Title Priority Date Filing Date
EP10185728A Withdrawn EP2357582A1 (fr) 1999-10-27 2000-10-27 Procédés et dispositifs pour identifier les motifs de systèmes biologiques et leurs procédés d'utilisation

Country Status (5)

Country Link
EP (2) EP1236173A2 (fr)
JP (1) JP5064625B2 (fr)
AU (1) AU779635B2 (fr)
CA (1) CA2388595C (fr)
WO (1) WO2001031580A2 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108346144A (zh) * 2018-01-30 2018-07-31 哈尔滨工业大学 基于计算机视觉的桥梁裂缝自动监测与识别方法
CN110189151A (zh) * 2019-06-12 2019-08-30 北京奇艺世纪科技有限公司 一种账号检测方法及相关设备

Families Citing this family (55)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1741036A (zh) 2000-06-19 2006-03-01 科雷洛吉克***公司 构造分类属于不同状态的生物样本的模型的方法
US6925389B2 (en) 2000-07-18 2005-08-02 Correlogic Systems, Inc., Process for discriminating between biological states based on hidden patterns from biological data
AU2002241535B2 (en) 2000-11-16 2006-05-18 Ciphergen Biosystems, Inc. Method for analyzing mass spectra
KR20030032395A (ko) * 2001-10-24 2003-04-26 김명호 서포트 벡터 머신을 이용한 다중 에스엔피(ｓｎｐ)와질병의 상관관계 분석 방법
AU2003268031A1 (en) 2002-07-29 2004-02-16 Correlogic Systems, Inc. Quality assurance/quality control for electrospray ionization processes
US9342657B2 (en) * 2003-03-24 2016-05-17 Nien-Chih Wei Methods for predicting an individual's clinical treatment outcome from sampling a group of patient's biological profiles
US6977370B1 (en) 2003-04-07 2005-12-20 Ciphergen Biosystems, Inc. Off-resonance mid-IR laser desorption ionization
WO2004094460A2 (fr) 2003-04-17 2004-11-04 Ciphergen Biosystems, Inc. Polypeptides lies aux peptides natriuretiques, procedes d'identification et d'utilisation de ces derniers
EP2369348A1 (fr) 2003-11-07 2011-09-28 Ciphergen Biosystems, Inc. Biomarqueurs pour la maladie d'Alzheimer
CA2547861A1 (fr) 2003-12-05 2005-06-23 Ciphergen Biosystems, Inc. Biomarqueurs seriques de la maladie de chagas
JP4774534B2 (ja) 2003-12-11 2011-09-14 アングーク ファーマシューティカル カンパニー，リミティド 集中化適応モデル及び遠隔操作サンプルプロセッシングの使用を介した生物学的状態の診断方法
WO2005098446A2 (fr) 2004-03-31 2005-10-20 The Johns Hopkins University Biomarqueurs du cancer des ovaires
JP2007535324A (ja) 2004-04-26 2007-12-06 チルドレンズ メディカル センター コーポレーション 疾患検出のための血小板バイオマーカー
US7811772B2 (en) 2005-01-06 2010-10-12 Eastern Virginia Medical School Apolipoprotein A-II isoform as a biomarker for prostate cancer
EP2993474B1 (fr) 2005-06-24 2019-06-12 Vermillion, Inc. Biomarqueurs pour le cancer de l'ovaire: la beta-2-microglobuline
EP2469279A1 (fr) 2006-03-11 2012-06-27 The Board Of Trustees Of The Leland Stanford Junior University Cystatin C, lysozyme et bêta-2 microglobuline en tant que biomarqueur d'une maladie des artères périphériques
JP5307996B2 (ja) * 2006-09-06 2013-10-02 株式会社Ｄｎａチップ研究所 判別因子セットを特定する方法、システム及びコンピュータソフトウェアプログラム
US8221984B2 (en) 2007-03-27 2012-07-17 Vermillion, Inc. Biomarkers for ovarian cancer
AU2008251381B2 (en) 2007-05-11 2014-10-30 The Johns Hopkins University Biomarkers for melanoma
JP2010532484A (ja) 2007-06-29 2010-10-07 コレロジック システムズ，インコーポレイテッド 卵巣癌のための予測マーカー
EP2220506B1 (fr) 2007-10-29 2013-10-02 Vermillion, Inc. Biomarqueurs permettant la détection du cancer des ovaires à un stade précoce
EP2252889B1 (fr) * 2008-02-08 2020-10-07 Health Discovery Corporation Procédé et système d analyse de données de cytométrie de flux utilisant des machines à vecteurs de support
JP5533662B2 (ja) * 2008-10-30 2014-06-25 コニカミノルタ株式会社 情報処理装置
WO2010060746A2 (fr) * 2008-11-26 2010-06-03 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. Procédé et dispositif d'analyse automatique de modèles
US8972899B2 (en) 2009-02-10 2015-03-03 Ayasdi, Inc. Systems and methods for visualization of data analysis
US9367812B2 (en) * 2010-08-25 2016-06-14 Optibrium Ltd. Compound selection in drug discovery
WO2012051519A2 (fr) 2010-10-14 2012-04-19 The Johns Hopkins University Biomarqueurs d'une lésion cérébrale
JPWO2012091093A1 (ja) * 2010-12-28 2014-06-05 啓 田代 緑内障診断チップと変形プロテオミクスクラスター解析による緑内障統合的判定方法
JP5672035B2 (ja) * 2011-02-03 2015-02-18 富士通株式会社 入力パラメータ算出方法、装置及びプログラム
WO2013003350A2 (fr) 2011-06-27 2013-01-03 Eisai R&D Management Co., Ltd. Arnmi comme biomarqueurs indicateurs de la maladie d'alzheimer
CN107674071B (zh) 2012-05-11 2021-12-31 同步制药公司 作为隐花色素调节剂的含有咔唑的磺酰胺类
WO2014071281A1 (fr) 2012-11-02 2014-05-08 The Johns Hopkins University Biomarqueurs de méthylation de l'adn pour évaluer le risque de dépression survenant après l'accouchement
EP4286853A3 (fr) 2013-05-10 2024-03-06 Johns Hopkins University Compositions présentant une grande spécificité pour l'évaluation du cancer de l'ovaire
CA2918054C (fr) 2013-07-11 2022-12-13 The Johns Hopkins University Biomarqueurs de type methylation de l'adn et genotype specifiques des tentatives de suicide et/ou des idees suicidaires
US10534003B2 (en) 2013-07-17 2020-01-14 The Johns Hopkins University Multi-protein biomarker assay for brain injury detection and outcome
WO2015120382A1 (fr) 2014-02-07 2015-08-13 The Johns Hopkins University Prédiction de réponse à un traitement utilisant des médicaments épigénétiques
TWI690521B (zh) 2014-04-07 2020-04-11 美商同步製藥公司 作為隱花色素調節劑之含有咔唑之醯胺類、胺基甲酸酯類及脲類
US10222386B2 (en) 2014-09-19 2019-03-05 The Johns Hopkins University Biomarkers of congnitive dysfunction
WO2016057485A1 (fr) 2014-10-06 2016-04-14 The Johns Hopkins University Biomarqueur spécifique de la méthylation d'adn et du génotype pour prédiction d'un état de stress post-traumatique
WO2016134365A1 (fr) 2015-02-20 2016-08-25 The Johns Hopkins University Biomarqueurs de blessure myocardique
US9988624B2 (en) 2015-12-07 2018-06-05 Zymergen Inc. Microbial strain improvement by a HTP genomic engineering platform
US11208649B2 (en) 2015-12-07 2021-12-28 Zymergen Inc. HTP genomic engineering platform
KR102006320B1 (ko) * 2015-12-07 2019-08-02 지머젠 인코포레이티드 Htp 게놈 공학 플랫폼에 의한 미생물 균주 개량
US10748277B2 (en) 2016-09-09 2020-08-18 Siemens Healthcare Gmbh Tissue characterization based on machine learning in medical imaging
WO2018163435A1 (fr) * 2017-03-10 2018-09-13 Omron Corporation Génération de données d'apprentissage
US10984334B2 (en) * 2017-05-04 2021-04-20 Viavi Solutions Inc. Endpoint detection in manufacturing process by near infrared spectroscopy and machine learning techniques
US20190034594A1 (en) * 2017-07-31 2019-01-31 National Cardiac, Inc. Computer-based systems and methods for monitoring the heart muscle of a patient with comprehensive contextual oversight
EP3732489A1 (fr) 2017-12-29 2020-11-04 Abbott Laboratories Nouveaux biomarqueurs et méthodes de diagnostic et d'évaluation d'une lésion cérébrale d'origine traumatique
US20210239700A1 (en) 2018-05-04 2021-08-05 Abbott Laboratories Hbv diagnostic, prognostic, and therapeutic methods and products
KR102245270B1 (ko) * 2019-02-25 2021-04-26 서강대학교 산학협력단 학습 데이터에 대한 오버샘플링 방법
WO2020172712A1 (fr) 2019-02-27 2020-09-03 Epiaxis Therapeutics Pty Ltd Procédés et agents pour évaluer une fonction de lymphocyte t et prédire une réponse à une thérapie
US20220093252A1 (en) * 2020-09-23 2022-03-24 Sanofi Machine learning systems and methods to diagnose rare diseases
WO2023122723A1 (fr) 2021-12-23 2023-06-29 The Broad Institute, Inc. Panels et procédés de diagnostic et de traitement du cancer du poumon
WO2024044578A1 (fr) 2022-08-22 2024-02-29 University Of Virginia Biomarqueurs de méthylation d'adn de trouble dysphorique prémenstruel et de dépression périménopausique
CN116582133B (zh) * 2023-07-12 2024-02-23 东莞市联睿光电科技有限公司 一种变压器生产过程数据智能管理***

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5143854A (en) 1989-06-07 1992-09-01 Affymax Technologies N.V. Large scale photolithographic solid phase synthesis of polypeptides and receptor binding screening thereof
US5837832A (en) 1993-06-25 1998-11-17 Affymetrix, Inc. Arrays of nucleic acid probes on biological chips
US5649068A (en) 1993-07-27 1997-07-15 Lucent Technologies Inc. Pattern recognition system using support vectors
WO1999022019A1 (fr) * 1997-10-29 1999-05-06 Rutgers, The State University Of New Jersey Etablissement d'un lien entre une sequence genique et une fonction genique par determination de la structure proteique tridimensionnelle
US7321828B2 (en) * 1998-04-13 2008-01-22 Isis Pharmaceuticals, Inc. System of components for preparing oligonucleotides
EP2296105B1 (fr) * 1998-05-01 2012-10-10 Health Discovery Corporation Traitement de données au moyen de machines à vecteurs de soutien

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See references of WO0131580A2 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108346144A (zh) * 2018-01-30 2018-07-31 哈尔滨工业大学 基于计算机视觉的桥梁裂缝自动监测与识别方法
CN108346144B (zh) * 2018-01-30 2021-03-16 哈尔滨工业大学 基于计算机视觉的桥梁裂缝自动监测与识别方法
CN110189151A (zh) * 2019-06-12 2019-08-30 北京奇艺世纪科技有限公司 一种账号检测方法及相关设备

Also Published As

Publication number Publication date
CA2388595C (fr) 2010-12-21
WO2001031580A3 (fr) 2002-07-11
AU1242701A (en) 2001-05-08
JP2003529131A (ja) 2003-09-30
EP2357582A1 (fr) 2011-08-17
WO2001031580A2 (fr) 2001-05-03
JP5064625B2 (ja) 2012-10-31
CA2388595A1 (fr) 2001-05-03
AU779635B2 (en) 2005-02-03

Similar Documents

Publication Publication Date Title
CA2388595C (fr) Procedes et dispositifs pouvant identifier des modeles dans des systemes biologiques, et procedes d'utilisation
US6789069B1 (en) Method for enhancing knowledge discovered from biological data using a learning machine
US6760715B1 (en) Enhancing biological knowledge discovery using multiples support vector machines
US7542959B2 (en) Feature selection method using support vector machine classifier
US7117188B2 (en) Methods of identifying patterns in biological systems and uses thereof
Peng A novel ensemble machine learning for robust microarray data classification
WO2001031579A2 (fr) Procedes et dispositifs permettant d'identifier des motifs dans des systemes biologiques et procedes d'utilisation correspondants
Rifkin et al. An analytical method for multiclass molecular cancer classification
KR100724104B1 (ko) 멀티플 지지벡터장치를 사용하여 멀티플 데이터세트로부터의 지식발견 강화방법
US20030225526A1 (en) Molecular cancer diagnosis using tumor gene expression signature
CA2435254C (fr) Procedes d'identification de motifs dans des systemes biologiques et utilisations desdits procedes
WO2009067655A2 (fr) Procédés de sélection de particularités par apprentissage local ; marqueurs de pronostic du cancer du sein et de la prostate
AU2002253879A1 (en) Methods of identifying patterns in biological systems and uses thereof
Aziz et al. A weighted-SNR feature selection from independent component subspace for nb classification of microarray data
Krishnapuram et al. Joint classifier and feature optimization for cancer diagnosis using gene expression data
Dong et al. The use of emerging patterns in the analysis of gene expression profiles for the diagnosis and understanding of diseases
Xu et al. Comparison of different classification methods for breast cancer subtypes prediction
Tamayo et al. Microarray Data Analysis: Cancer Genomics and Molecular Pattern Recognition
Nilsson Nonlinear dimensionality reduction of gene expression data
Huiqing Effective use of data mining technologies on biological and clinical data
Aarthi et al. Enhancing sample classification for microarray datasets using genetic algorithm
Sivaraksa et al. Predictive gene lists for breast cancer prognosis: a topographic visualisation study
Orduña Cabrera et al. Bioinformatics: a promising field for case-based reasoning
Ma Effective techniques for gene expression data mining
Dinger Cluster Analysis of Gene Expression Data on Cancerous Tissue Samples

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20020515

AK Designated contracting states

Kind code of ref document: A2

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE

AX Request for extension of the european patent

Free format text: AL;LT;LV;MK;RO;SI

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: MCKENZIE, JOE

Owner name: CARLS, GARRY L.

Owner name: BERGERON, GLYNN

Owner name: O'HAYER, TIMOTHY P.

Owner name: SIMPSON, K. RUSSELL

Owner name: MATTHEWS, JOHN E.

Owner name: ANDERSON, CURTIS

Owner name: FARLEY, PETER J.

Owner name: PADEREWSKI, JULES B.

Owner name: ROBERTS, JAMES

Owner name: STERN, JULIAN N.

Owner name: MEMORIAL HEALTH TRUST, INC.

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: HEALTH DISCOVERY CORPORATION

17Q First examination report despatched

Effective date: 20070919

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: HEALTH DISCOVERY CORPORATION

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN

18D Application deemed to be withdrawn

Effective date: 20140408