CN110964800B

CN110964800B - cfRNA markers for predicting risk of preterm birth

Info

Publication number: CN110964800B
Application number: CN201910890694.3A
Authority: CN
Inventors: 张永彪; 马翠; 毛轲; 石小峰; 徐晓鹏; 尚策; 付博
Original assignee: Beihang University
Current assignee: Beihang University
Priority date: 2019-09-20
Filing date: 2019-09-20
Publication date: 2021-10-22
Anticipated expiration: 2039-09-20
Also published as: CN110964800A

Abstract

The invention discloses a cfRNA marker for predicting the risk of preterm birth, wherein the indicated cfRNA is selected from one or more of CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX 1. The invention also discloses a risk prediction model for predicting premature delivery, which is used as an auxiliary means for predicting the premature delivery risk of the pregnant woman, thereby carrying out risk assessment and detection on the patient.

Description

cfRNA markers for predicting risk of preterm birth

Technical Field

The invention belongs to the field of biomedicine, and relates to a cfRNA marker for predicting premature delivery risk.

Background

Preterm birth is a leading cause of morbidity and mortality in newborns worldwide and poses a significant health burden. About 1500 million premature babies are born each year worldwide, accounting for 5-15% of the total number of births, the incidence of premature birth in the united states is around 12%, and the incidence of premature birth in china is between 8% and 15%, with the highest incidence in guangzhou being up to 15%. The birth population of China is nearly 1600 million per year, and the birth rate is 12.29 percent, and more than one million premature infants are estimated to be born per year. It was found that among the important causes of perinatal mortality, preterm birth accounted for 80% at 24 weeks, 10% at 30 weeks, and a significant decrease was observed at 34 weeks. 70% of neonatal deaths, 30% of infant deaths, 25-50% of chronic nervous system injuries occur in premature infants. Early prediction and diagnosis, identification of true preterm high-risk pregnant women is critical to reducing preterm birth and associated complications, while excluding pregnant women at low risk of preterm birth in the short term is critical to reducing over-treatment. Due to the limited ability of fetal fibronectin (fn) and Cervical Length (CL) approaches to predict preterm birth, further studies are needed to explore other possible screening approaches.

Free nucleic acids are also called extracellular nucleic acids, and are extracellular free DNA (cfDNA) and RNA (cfRNA) widely present in body fluids such as plasma, saliva, alveolar lavage fluid, urine, semen, pleural effusion and the like, and in cell culture fluid. With the development of modern molecular biology technology, free nucleotides are widely studied and become important molecular markers for noninvasive diagnosis of diseases. Researches show that the plasma of pregnant women contains fetal cfDNA and cfRNA, so that noninvasive prenatal diagnosis technology is rapidly developed, and the risks of abortion and teratogenesis caused by traumatic prenatal diagnosis are avoided.

Multiple studies have demonstrated that cfRNA plays an important role in the field of preterm delivery, but up to now, noninvasive prenatal diagnostic techniques based on cfRNA are still blank in the prediction of preterm delivery in chinese populations. Due to the enormous number of cfrnas, research in the field of preterm delivery for cfrnas is still in the infancy, and the mechanism by which cfrnas influence the occurrence of preterm delivery is awaited for further research. Since the role of cfRNA in the mechanism of development of preterm birth has not yet been elucidated, it is of great significance to search for the molecular basis of cfRNA preterm birth.

Disclosure of Invention

In order to remedy the deficiencies of the prior art, it is an object of the present invention to provide a means and a product for predicting the risk of preterm birth.

The second objective of the present invention is to provide a method for establishing a preterm birth prediction model.

In order to achieve the purpose, the invention adopts the following technical scheme:

the invention provides application of cfRNA in preparing a product for predicting the risk of premature delivery, wherein the cfRNA is selected from one or more of CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX 1.

Further, the product comprises a chip, or a kit. Wherein the chip comprises a gene chip; the kit comprises a gene detection kit. The gene chip comprises a solid phase carrier and oligonucleotide probes fixed on the solid phase carrier, wherein the oligonucleotide probes comprise oligonucleotide probes aiming at CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1 genes for detecting the gene expression level; the gene detection kit comprises primers or chips for detecting the expression levels of CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1 genes.

Further, the product comprises an agent that detects the expression level of cfRNA (CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, and RBX1) in sample plasma.

Further, the agent is selected from:

probes that specifically recognize CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, and RBX 1; or

Primers that specifically amplified CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, and RBX 1.

The present invention provides a product for predicting the risk of preterm birth by measuring the levels of cfRNA (CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1) in plasma.

Further, the products include (but are not limited to): expression levels of cfRNA (CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, and RBX1) genes in plasma were detected by real-time quantitative PCR, chip, or high throughput sequencing platform.

Wherein, the product for predicting premature delivery by real-time quantitative PCR at least comprises a pair of primers for specifically amplifying CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1 genes; the product for predicting premature birth by the chip comprises probes hybridized with nucleic acid sequences of CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1 genes.

Further, after each sample is detected respectively, predicting the sample to be classified by using a random forest model and a predict function, wherein the use function is as follows: predict (random forest _ model, data _ unknown _ sample _ data), where random forest _ model is a random forest model and unknown _ sample _ data is the expression level of each cfRNA in a sample to be classified.

Further, when the individual premature delivery risk is evaluated, the cfRNA expression level of a sample to be tested is put into the function, so that the judgment result of the random forest model is obtained, the result is expressed by 1 or 2, 1 is expressed as a normal pregnant woman, and 2 is expressed as the pregnant woman having the premature delivery risk.

The invention provides a method for establishing a preterm delivery prediction model, which comprises the following steps:

(1) dividing pregnant women into a training group and a verification group, and measuring the differential expression cfRNA in a specimen collected by a subject in the training group;

(2) screening characteristic genes in the differential expression cfRNA based on two algorithms of LASSO and Boruta;

(3) establishing a risk prediction model based on a random forest method;

(4) and (5) verifying the risk prediction model by using a verification group, and checking the prediction accuracy of the established model.

Further, the feature variables include CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, and RBX 1.

In the invention, one or more selected from CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1 are subjected to model construction, and the result shows that one or more have the capacity of predicting premature delivery, but the combined prediction of premature delivery by 13 cfRNAs has stronger judgment efficiency.

Further, the risk prediction model is random forest _ model, the used function is predict (random forest _ model, data is unknown _ sample _ data), whether the risk of premature birth exists is judged according to the function result, and when the judgment result is 1, the pregnant woman is normal; when the result is 2, it indicates that the pregnant woman is at risk of preterm birth.

The "sample" is plasma.

In the present invention, the term "probe" refers to a molecule that binds to a specific sequence or subsequence or other portion of another molecule. Unless otherwise indicated, the term "probe" generally refers to a polynucleotide probe that is capable of binding to another polynucleotide (often referred to as a "target polynucleotide") by complementary base pairing. Depending on the stringency of the hybridization conditions, a probe can bind to a target polynucleotide that lacks complete sequence complementarity to the probe. The probe may be directly or indirectly labeled, and includes within its scope a primer. Hybridization modalities, including, but not limited to: solution phase, solid phase, mixed phase or in situ hybridization assays.

Exemplary probes in the present invention include PCR primers as well as gene-specific DNA oligonucleotide probes, such as microarray probes immobilized on a microarray substrate, quantitative nuclease protection test probes, probes attached to molecular barcodes, and probes immobilized on beads.

In the present invention, the "kit" further contains a label for labeling the RNA sample, and a substrate corresponding to the label. In addition, the kit may further include various reagents required for RNA extraction, PCR, hybridization, color development, and the like, including but not limited to: an extraction solution, an amplification solution, a hybridization solution, an enzyme, a control solution, a color development solution, a washing solution, and the like. In addition, the kit also comprises an instruction manual and/or chip image analysis software.

The term "differentially expressed cfRNA" or "cfRNA differentially expressed" refers to cfRNA whose expression is significantly different at cfRNA levels of one gene under different environmental stresses, times, spaces, and the like. Structural changes such as gene mutation or methylation occur under different factors, resulting in different cfrnas.

The method for establishing the premature delivery risk prediction model is a random forest, the essence of a random forest algorithm is a classifier integration algorithm based on decision trees, each tree depends on a random vector, and all vectors of the random forest are independently and identically distributed. The random forest is to randomize the column variables and row observations of the data set to generate a plurality of classification numbers, and finally summarize the classification tree results. Compared with a neural network, the random forest reduces the operation amount and improves the prediction precision, and the algorithm is insensitive to multivariate collinearity and is more stable to missing data and unbalanced data, so that the algorithm can be well adapted to thousands of interpretation variable data sets.

The invention has the advantages that:

the invention discovers for the first time that the combined detection of cfRNA (CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1) in plasma can effectively predict the preterm birth risk of pregnant women.

The invention provides a model for predicting preterm delivery, which can predict the risk of preterm delivery of a pregnant woman by detecting the expression level of cfRNA (CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D and RBX1) in the plasma of the pregnant woman and substituting the expression level into a preterm delivery prediction model, thereby carrying out risk assessment and detection on the pregnant woman.

Drawings

FIG. 1 is a result of importance of an import function computation model variable;

fig. 2 is a graphical representation of AUC values and accuracy for single characteristic value determination of preterm delivery and for multiple characteristic values combined determination of preterm delivery.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and examples. The following examples are intended to illustrate the invention only and are not intended to limit the scope of the invention. The experimental procedures for the specific conditions not indicated in the examples are generally carried out according to conventional conditions, for example Sambrook et al, molecular cloning: the conditions described in the laboratory Manual (New York: Cold Spring harbor laboratory Press,1989), or according to the manufacturer's recommendations.

Example 1 screening of Gene markers associated with preterm birth

1. Sample(s)

208 samples with complete clinical information are screened from samples stored in hospitals, wherein 156 normal pregnant women and 52 pregnant women who have preterm birth are randomly divided into a training group (156) and a verification group (52), and the clinical information of the included patients comprises the age of the pregnant women, BMI (BMI value), pregnancy history, interval between two pregnancies, drinking history, smoking history, drug absorption history, late abortion and/or preterm birth history, cervical operation history, vaginal ultrasonic examination results, abnormal fetal and amniotic fluid quantity, multiple pregnancy, pregnancy complications or complications of pregnancy, assisted pregnancy by assisted reproductive technology, production mode, term production, difficult labor and the like.

2. Data normalization

Raw cfRNA expression data was collected and normalized to facilitate comparison between samples, with the formula:

3. differential expression cfRNA analysis

The gene differential expression analysis was performed using the exact test, glmLRT and glmQLFTest functions in the edgeR package in the R language with a screening criterion of PValue < 0.01.

4. Results

The common differential expression genes obtained by the three functions of exattest, glmLRT and glmQLFTest are the final result, and 68 differential expression cfrnas are obtained.

Example 2 feature selection procedure

1. Selection of important feature variables using LASSO algorithm

The loss function of LASSO is minimized as follows:

and (3) performing loss function solution based on the glmnet packet in the R language, and simply selecting different lambda values to obtain different w. By selecting the optimal parameters, the error rate is minimized. And (3) selecting characteristic variables by using LASSO, and screening 16 cfRNAs.

2. Selection of important feature variables using the Boruta algorithm

The Boruta algorithm is a packing algorithm around a random forest. When fitting a random forest model to a data set, you can recursively handle underperforming features in each iteration. The method reduces the error of the random forest model to the maximum extent, and the minimum optimal feature subset is formed finally.

The boruta algorithm runs as follows:

1) first, it adds randomness to a given data set by creating all of the features of the mixed copy (i.e., shadow features).

2) It then trains an extended dataset of random forest classifications and takes a feature importance measure (default setting is average reduction accuracy) to evaluate the importance of each feature, higher meaning more important.

3) In each iteration it checks whether a real feature is of higher importance than the best shadow feature (i.e. whether the feature scores higher than the largest shadow feature) and continuously deletes features that it considers to be very unimportant.

4) Finally, the algorithm stops when all features are confirmed or rejected, or the algorithm reaches a specified limit for random forest runs.

Feature variable selection was performed based on the Boruta package in the R language, and 28 cfRNAs were screened.

3. Results

The genes screened based on the two LASSO and Boruta algorithms are intersected to obtain 13 common differentially expressed cfrnas: CLCN3, DAPP1, PPBP, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, and RBX 1.

Example 3 prediction of risk of preterm labor in pregnant women by random forest model

Estimation process of random forest

1) The value of m is designated, namely m variables are randomly generated to be used for a binary tree on the nodes, the selection of the variables of the binary tree still meets the principle that the purity of the nodes is minimum, and m in the model is 2;

2) randomly extracting k sample sets in a place-back manner in an original data set by applying a Bootstrap self-service method to form k decision trees, and for the prediction of a single decision tree by using samples which are not extracted, when k is 300 in the model, the model is basically stable;

3) predicting an unknown _ sample to be classified according to a random forest model random forest _ model consisting of k decision trees, wherein a used function is a prediction (random forest _ model, data is unknown _ sample _ data), and the prediction principle is simple average;

4) calculating the importance of the model variables by using the import function, and determining which variables have the largest contribution to the model, wherein the result is shown in FIG. 1;

5) the ROC curve judges the accuracy of judging the premature delivery by judging a single characteristic value and jointly judging the premature delivery by a plurality of characteristic values, and the results are shown in Table 1. With the increase of the characteristic genes, the area under the curve (AUC) and the accuracy of judgment gradually increased, and the AUC and the accuracy formed after the combination of 13 cfrnas were the highest, and the results are shown in fig. 2.

TABLE 1 area under the curve formed by different cfRNAs and accuracy of the determination

Claims

1. Use of an agent that detects the expression level of cfRNA in the manufacture of a product for predicting the risk of preterm birth, wherein the cfRNA is DAPP1, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, RBX1, PPBP and CLCN 3.

2. The use of claim 1, wherein the product comprises a chip, or a kit.

3. The use according to claim 1, wherein the agent is selected from the group consisting of: probes that specifically recognize DAPP1, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, RBX1, PPBP, and CLCN 3;

or primers that specifically amplify DAPP1, MAP3K7CL, MOB1B, RAB27B, RGS18, TTC33, RPA1, RAP1B, HMGN3, ATP5D, RBX1, PPBP, and CLCN 3.