Differential gene identification method based on nonparametric test

Through the method based on the maximum mean difference (MMD) and random messy rearrangement of kernel functions, the problem of differential expression gene detection in time series RNA-Seq data in the prior art is solved, and the time points and repetition are achieved. Effective analysis of low data improves the flexibility and accuracy of analysis.

CN118918949AActive Publication Date: 2024-11-08WUHAN BOTANICAL GARDEN CHINESE ACAD OF SCI
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202410967525.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-08
Estimated Expiration
2044-07-18

AI Technical Summary

Technical Problem

Prior art When processing time series RNA-Seq data, it is difficult to reliably detect differentially expressed genes from data sets with very little time point and repetitiveness, and there is a lack of flexible differential expression analysis methods to process different types of experimental data.

Method used

Using a non-parametric statistical test based on kernel function, namely the maximum mean difference (MMD), the significance parameter Padj of the gene is calculated by calculating unbiased empirical estimates between gene expression level sequences, and combined with the method of random disruption rearrangement, the significance parameter of the gene is calculated. to identify differentially expressed genes.

Benefits of technology

This method can more flexibly process data with sparse time points or large fluctuations in gene expression levels, capture nonlinear dynamic changes in gene expression, improve the flexibility and applicability of analysis, especially in the case of small number of time points, which can be effective Identify differential genes and maintain low false positive rates.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118918949A_ABST
    Figure CN118918949A_ABST
Patent Text Reader

Abstract

The invention discloses a differential gene identification method based on nonparametric test. The method comprises the following steps: acquiring gene expression matrixes of an experimental group and a control group; calculating MMD unbiased empirical estimation between the gene expression level sequence of the gene in the experimental group and the gene expression level sequence of the gene in the control group as original MMD unbiased empirical estimation of the gene; after the gene expression level sequence of the gene in the experimental group and the gene expression level sequence of the gene in the control group are combined and then randomly disrupted and rearranged, MMD unbiased empirical estimation is calculated as new MMD unbiased empirical estimation, significance parameters of the gene are calculated, and differential expression genes are judged according to the significance parameters. The method does not depend on specific data distribution hypothesis, and a data set with rare time points or large gene expression level fluctuation can be processed more flexibly. The method is suitable for various types of gene expression data, and the flexibility and applicability of analysis are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of bioinformatics, and specifically relates to a differential gene identification method based on non-parametric testing, which is suitable for gene identification. Background Art

[0002] Since gene expression regulation is a dynamic process, characterizing the genome-wide dynamic regulation of gene expression is essential for understanding transcriptome changes in biological processes. Time-series RNA-Seq data enable us to study the dynamic behavior of gene expression by collecting gene expression levels from biological samples of the same subject at different time points. With the advantage of high-throughput transcriptomics, the acquisition of time-series RNA-Seq data has increased significantly, making time-series gene expression experiments increasingly common. The main goal of such experiments is to identify genes that are differentially expressed between two conditions. This process is crucial because detecting differentially expressed genes through time-series analysis not only promotes a deeper understanding of gene regulatory mechanisms under temporally varying conditions and stimuli, but also provides the possibility of discovering new biomarkers. Due to the high cost of experiments or the limited number of unique individual biological samples, time-series RNA-Seq datasets usually contain few time points and biological replicates. However, the basic scientific need to identify differentially expressed genes associated with biological processes has not diminished. Therefore, there is an urgent need to reliably detect differentially expressed genes from datasets with few time points and replicates. In addition, due to the diversity of experimental designs that generate different types of data, flexible differential expression analysis methods are urgently needed to process data from any experiment.

[0003] Several differential expression analysis methods have been developed for time series RNA-seq datasets. For example, NextmaSigPro-GLM is designed to identify significant differential expression profiles across multiple experimental time points or conditions in RNA-seq data, and the model utilizes a negative binomial generalized linear model to accommodate the complexity and dynamics of gene expression over time. splineTC is an R package specifically designed for time series analysis. It utilizes spline functions to smoothly model and analyze the temporal trends of data. This approach is particularly valuable for understanding dynamic patterns of gene expression over time. ImpulseDE2 uses a unique algorithm that models gene expression patterns as impulse-like responses, making it good at detecting transient or dynamic changes in gene expression. Both DESeq2 and edgeR are based on the negative binomial model and are considered the gold standard in the field of differential expression analysis. Although traditional comparisons using these two tools do not take into account the analysis of time series data, their sophisticated designs allow for simple investigations of time series data. Summary of the invention

[0004] The purpose of the present invention is to address the above-mentioned problems existing in the prior art and provide a differential gene identification method based on non-parametric test. By applying a non-parametric statistical test based on a kernel function, namely, the maximum mean difference (MMD), the differentially expressed genes in time series RNA-Seq data are identified and analyzed.

[0005] The above-mentioned purpose of the present invention is achieved by the following technical means:

[0006] The differential gene identification method based on non-parametric test includes the following steps:

[0007] Step 1: Obtain the gene expression matrix of the experimental group and the control group. Each column of the gene expression matrix represents a time point, and each element of each row of the gene expression matrix constitutes the gene expression level sequence of the gene corresponding to the row;

[0008] Step 2: Calculate the MMD unbiased empirical estimate between the gene expression level sequence of gene i in the experimental group and the gene expression level sequence in the control group as the original MMD unbiased empirical estimate of gene i;

[0009] Step 3: The gene expression levels of the gene expression level sequence of gene i in the experimental group and the gene expression level sequence of gene i in the control group are combined and randomly shuffled to generate a new gene expression level sequence of gene i in the experimental group and a new gene expression level sequence in the control group, and the MMD unbiased empirical estimate is calculated as a new MMD unbiased empirical estimate. The significance parameter P of gene i is calculated based on the original MMD unbiased empirical estimate and the new MMD unbiased empirical estimate. adj , if the significance parameter P of gene i adj If the value is less than the significance threshold, gene i is considered to be a differentially expressed gene.

[0010] As described above, obtaining the gene expression matrix of the experimental group and the control group in step 1 includes the following steps: collecting RNA-seq raw data of the experimental group and the control group under the same temperature environment and at multiple time points, and obtaining the gene expression matrix of the experimental group and the control group respectively after data preprocessing.

[0011] The unbiased empirical estimate of the MMD in steps 2 and 3 as described above is calculated based on the following formula:

[0012]

[0013] Among them, MMD 2 (X i ,Y i ) is the unbiased empirical estimate of MMD of gene i, X i is the gene expression level sequence of gene i at m time points in the experimental group, is the gene expression level of gene i from the 1st time point to the mth time point in the gene expression matrix of the experimental group, Y i is the gene expression level sequence of gene i at n time points in the control group, is the gene expression level of gene i in the gene expression matrix of the control group from the 1st time point to the nth time point, m is the total number of time points in the gene expression matrix of the experimental group, n is the total number of time points in the gene expression matrix of the control group, iu1 and iu2 are the ordinal variables of the time points of the experimental group, iv1 and iv2 are the ordinal variables of the time points of the control group, and k is the inner product function.

[0014] As mentioned above, the significance parameter P of gene i adj Calculated based on the following formula:

[0015]

[0016] Where r is the number of random shuffles, R is the total number of random shuffles, I(·) is the indicator function, and the gene expression levels of the gene expression level sequence of gene i in the experimental group and the gene expression level sequence in the control group are combined and randomly shuffled: the new gene expression level sequence of the experimental group is X ir , the gene expression level sequence of the new control group is Y ir , the new unbiased empirical estimate of MMD is calculated as 2 [X ir ,Y ir ].

[0017] As described above, the content in the brackets of the indicator function represents a condition. When the condition is met, the value of the indicator function is 1, and when the condition is not met, the value of the indicator function is 0.

[0018] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned differential gene identification method when executing the computer program.

[0019] A computer-readable storage medium stores a computer program, which implements the steps of the above-mentioned differential gene identification method when executed by a processor.

[0020] A computer program product includes a computer program, which implements the steps of the above-mentioned differential gene identification method when executed by a processor.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] This method does not rely on specific data distribution assumptions and can more flexibly handle data sets with sparse time points or large fluctuations in gene expression levels. This method can better capture the nonlinear dynamic changes of gene expression levels in experimental and control groups through MMD unbiased empirical estimation. According to the heterogeneity and complexity of gene expression data, different types of kernel functions (such as linear, polynomial, radial basis function) can be selected to make this method applicable to various types of gene expression data and improve the flexibility and applicability of analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flow chart of the present invention;

[0024] Figure 2 is a comparative data graph of an embodiment of the present invention. The performance of different methods in gene expression data analysis is shown by drawing a receiver operating characteristic curve (ROC curve). KMMDE is the method of this paper. Gene expression data with different time points and without biological replication were used for comparative analysis. When making comparisons, the other methods shown in the figure (edgeR, DESeq2, maSigPro, splineTC and Funpat) also calculated the adjusted P adj value to control the false discovery rate. The numbers in the bar above the picture represent that the data has 4 time points, 6 time points, 8 time points, and 10 time points. The numbers in the right column represent the number of genes with different differences. The horizontal axis (False Positive Rate) represents the false positive rate, that is, the proportion of the true null hypothesis (genes with no difference) being mistakenly judged as having a difference. The vertical axis (True Positive Rate) represents the true positive rate, that is, the proportion of the true non-null hypothesis (genes with differences) being correctly judged as having a difference. As can be seen from the curve, the KMMDE method (red curve) is above other methods in most cases, especially when the number of time points is small, it still shows a certain advantage in the higher true positive rate area (the curve is close to the upper left corner). This shows that KMMDE can more effectively identify genes that are truly different while maintaining a low false positive rate. At the same time, KMMDE shows good stability in scenarios with different time points and the number of differential genes, which is very important when facing the complexity of actual biological data. DETAILED DESCRIPTION

[0025] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below with reference to examples. The implementation examples described herein are only used to illustrate and explain the present invention and are not intended to limit the present invention.

[0026] Example:

[0027] The differentially expressed gene identification method based on non-parametric test takes the temporal expression data of wild grape (experimental group) and common grape (control group) under the same cold environment as an example. Figure 1 As shown, the following steps are included:

[0028] Step 1: Obtain the gene expression matrix. Collect the original RNA-seq data of the experimental group and the control group at the same temperature (e.g., 5-15 degrees) and at multiple time points, and obtain the gene expression matrices of the experimental group and the control group after data preprocessing. Each column of the gene expression matrix represents a time point, and each row represents a gene. The elements in the gene expression matrix represent the gene expression level of the gene corresponding to the row at the specific time point corresponding to the column. Each element in each row of the gene expression matrix constitutes the gene expression level sequence of the gene corresponding to the row.

[0029] Step 2: Calculate the maximum mean difference of gene i. Assume that the gene expression matrix of the experimental group has m time points and the gene expression matrix of the control group has n time points. i ) is the sequence of gene expression levels at m time points: is the gene expression level of gene i from the 1st time point to the mth time point in the gene expression matrix of the experimental group, and gene i in the control group (Y i ) is the gene expression level sequence at n time points in i ={y i1 ,y i2 ,...y in}, is the gene expression level of gene i from the 1st time point to the nth time point in the gene expression matrix of the control group. The gene expression level sequence X i and gene expression level sequence Y i Mapped into a high-dimensional feature space. Calculate the gene expression level sequence X i and gene expression level sequence Y i The inner product k(X i ,Y i ):

[0030] k(X i ,Y i )=<φ(X i ),φ(Y i )> F (1)

[0031] Where k is the inner product function, <,> F is the inner product operator, and F represents the Frobenius norm.

[0032] The gene expression level sequence X with a sample size of mi With a sample size of n gene expression level sequence Y i The unbiased empirical estimate of the MMD between (i.e., the original unbiased empirical estimate of the MMD of gene i) is:

[0033]

[0034] Among them, k is the inner product function, iu1, iu2 are the ordinal variables of the experimental group at the time point, ranging from 1 to m; iv1, iv2 are the ordinal variables of the control group at the time point, ranging from 1 to n. Among them, iu1≠iu2 means that the case of iu1=iu2 is excluded in the summation process, and iv1≠iv2 and iu1≠iv2 are the same.

[0035] Step 3: Significance analysis, calculate the adjusted significance parameter P of gene i adj In order to calculate the significance of the difference in expression levels of gene i between the experimental group and the control group, we used the permutation test to evaluate whether the original MMD value was significant. First, the gene expression levels of the experimental group and the control group of gene i were merged into one sequence as the merged sequence, and then the merged sequence was randomly shuffled R times, that is, {x i1 ,x i2 ,...x im ,y i1 ,y i2 ,...y in} Rearrange R times to generate R After each rearrangement, the first m gene expression levels are selected from the rearranged sequence as the gene expression level sequence X of the new experimental group. ir , select the last n gene expression levels as the gene expression level sequence Y of the new control group ir The gene expression level sequence of the new experimental group and the new control group corresponding to the gene i obtained by the r-th random shuffle are subjected to MMD analysis based on formula 2 to obtain the new unbiased empirical estimate of MMD corresponding to gene i. The gene expression level sequence of the new experimental group after the r-th random shuffle is recorded as X ir , the gene expression level sequence of the new control group is Y ir , the new unbiased empirical estimate of MMD is calculated as 2 [X ir ,Y ir ]. By rearranging R times, we can generate a distribution of MMD unbiased empirical estimates, which is used to evaluate the significance of the original MMD unbiased empirical estimate. The adjusted significance parameter P adjThe significance parameter P for gene i is calculated by comparing the original unbiased empirical estimate of MMD with the new unbiased empirical estimate of MMD generated by the rearrangement to determine whether the difference in gene expression level is statistically significant. adj It can be defined as:

[0036]

[0037] Where r is the number of random shuffles, R is the total number of random shuffles, formula I(·) is the indicator function, and the content in brackets represents the condition. When the condition is true (condition is established), the value of the indicator function is 1, and when the condition is false (condition is not established), the indicator function is 0. Accumulate all the MMD 2 [X ir ,Y ir ]>MMD 2 [X i ,Y i ] case. In order to increase the robustness of the calculation and avoid P adj If the result is 0, add 1 to the accumulated result and then divide it by R + 1. This method ensures that the error rate is effectively controlled in multiple comparisons, making the results more reliable.

[0038] Get the significance parameter P of gene i adj After that, the significance threshold is set to 0.05. If the significance parameter P of gene i is adj <0.05, gene i is considered to be a differentially expressed gene, which means that there is a statistically significant difference in the expression levels of these genes between the experimental group and the control group.

[0039] The following is a comparison result between the method of the present invention and the current mainstream tools. Among them, KMMDE is the method of the present invention. Gene expression level sequences with different numbers of time points and without biological replication were used for comparative analysis. The numbers in the column above the picture represent that the gene expression level sequence has 4 time points, 6 time points, 8 time points, and 10 time points. The numbers in the right column represent the gene expression level sequences with different numbers of differential genes. The horizontal axis (FalsePositive Rate) represents the false positive rate, that is, the proportion of the true null hypothesis (genes with no difference) being mistakenly judged as having a difference. The vertical axis (TruePositive Rate) represents the true positive rate, that is, the proportion of correctly judging the true non-null hypothesis (genes with differences) as having a difference. It can be seen from the curve that the KMMDE method (red curve) is above other methods in most cases, especially in the case of a small number of time points, and still shows certain advantages in the higher true positive rate area (the curve is close to the upper left corner). This shows that KMMDE can more effectively identify genes that are truly different while maintaining a low false positive rate. At the same time, KMMDE shows good stability at different time points and in different gene numbers, which is very important when dealing with the complexity of actual biological data.

[0040] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0041] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0042] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0043] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing related hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods.

[0044] It should be noted that the specific embodiments described in the present invention are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the attached claims.

Claims

1. A differential gene identification method based on non-parametric test, characterized in that: The following steps are involved: Step 1: Obtain the gene expression matrix of the experimental group and the control group. Each column of the gene expression matrix represents a time point, and each element of each row of the gene expression matrix constitutes the gene expression level sequence of the gene corresponding to the row; Step 2: Calculate the MMD unbiased empirical estimate between the gene expression level sequence of gene i in the experimental group and the gene expression level sequence in the control group as the original MMD unbiased empirical estimate of gene i; Step 3: The gene expression levels of the gene expression level sequence of gene i in the experimental group and the gene expression level sequence of gene i in the control group are combined and randomly shuffled to generate a new gene expression level sequence of gene i in the experimental group and a new gene expression level sequence in the control group, and the MMD unbiased empirical estimate is calculated as a new MMD unbiased empirical estimate. The significance parameter P of gene i is calculated based on the original MMD unbiased empirical estimate and the new MMD unbiased empirical estimate. adj , if the significance parameter P of gene i adj If the value is less than the significance threshold, gene i is considered to be a differentially expressed gene.

2. The differential gene identification method based on nonparametric test according to claim 1, characterized in that: Obtaining the gene expression matrices of the experimental group and the control group in step 1 includes the following steps: collecting RNA-seq raw data of the experimental group and the control group under the same temperature environment and at multiple time points, and obtaining the gene expression matrices of the experimental group and the control group respectively through data preprocessing.

3. The differential gene identification method based on non-parametric test according to claim 1, characterized in that: The unbiased empirical estimate of MMD in steps 2 and 3 is calculated based on the following formula: Among them, MMD 2 (X i , Y i ) is the unbiased empirical estimate of MMD of gene i, X i is the gene expression level sequence of gene i at m time points in the experimental group, is the gene expression level of gene i from the 1st time point to the mth time point in the gene expression matrix of the experimental group, Y i is the gene expression level sequence of gene i at n time points in the control group, is the gene expression level of gene i in the gene expression matrix of the control group from the 1st time point to the nth time point, m is the total number of time points in the gene expression matrix of the experimental group, n is the total number of time points in the gene expression matrix of the control group, iu1 and iu2 are the ordinal variables of the time points of the experimental group, iv1 and iv2 are the ordinal variables of the time points of the control group, and k is the inner product function.

4. The differential gene identification method based on non-parametric test according to claim 3, characterized in that: The significance parameter P of the gene i adj Calculated based on the following formula: Where r is the number of random shuffles, R is the total number of random shuffles, I(·) is the indicator function, and the gene expression levels of the gene expression level sequence of gene i in the experimental group and the gene expression level sequence in the control group are combined and randomly shuffled: the new gene expression level sequence of the experimental group is X ir , the gene expression level sequence of the new control group is Y ir , the new unbiased empirical estimate of MMD is calculated as 2 [X ir , Y ir ].

5. The differential gene identification method based on non-parametric test according to claim 4, characterized in that: The content in the brackets of the indicator function represents a condition. When the condition is met, the value of the indicator function is 1, and when the condition is not met, the value of the indicator function is 0.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the differential gene identification method described in any one of claims 1 to 5 are implemented.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the differential gene identification method according to any one of claims 1 to 5 are implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the differential gene identification method described in any one of claims 1 to 5 are implemented.

Citation Information

Patent Citations

  • Method for constructing signaling transduction network based on multivariate Granger test

    CN105825077A

  • Power distribution network small current grounding fault section positioning method based on maximum mean value difference

    CN113447761A

  • Lithium battery remaining service life prediction based on ACNN-Mogrimer LSTM-MMD

    CN117289135A

  • Method and system for analysis of gene-expression data

    US20060190190A1