A disease-related miRNAs prediction method and system based on deep autoencoder

Abstract

The embodiment of the specification provides a disease-related miRNA prediction method and system based on a deep autoencoder, wherein the method comprises the following steps: obtaining miRNA expression profile data of samples from a TCGA database, and performing standardization processing on the obtained miRNA expression profile data respectively; the samples comprise disease samples and normal samples; filling zero values in the miRNA expression profile data after the standardization processing by using a matrix decomposition method, and training a deep autoencoder model by using the filled miRNA expression profile data; training a reverse encoder based on the deep autoencoder model, obtaining the miRNA expression profile data of the disease samples and the disease correlation size according to the reverse encoder, and obtaining miRNAs with high disease correlation. The application can discover new disease-related miRNAs, and promote the diagnosis and treatment of diseases.

Description

TECHNICAL FIELD

[0001] The present document relates to the technical field of bioinformatics, and in particular to a disease-related miRNAs prediction method and system based on a deep autoencoder. BACKGROUND

[0002] According to the latest global cancer burden estimation data released by the International Agency for Research on Cancer (IARC) of the World Health Organization in 2020, there were 19.29 million new cancer cases worldwide in 2020, and 9.96 million cancer deaths. In China, cancer has become one of the common high-incidence diseases, and developing new cancer classification methods and improving cancer classification accuracy is an important research content. miRNAs are a class of non-coding short sequence molecules with a length of about 20 nt. Studies have shown that the abnormal expression of some miRNAs is closely related to the occurrence and deterioration of various complex diseases including cancer. With the development of sequencing technology, it has become convenient to obtain the expression information of micro miRNAs molecules in biological tissue samples, which provides data support for us to develop cancer classification methods based on miRNAs expression profiles.

[0003] miRNAs are a class of non-coding short sequence RNA molecules (about 20 nt in length). Although miRNAs cannot directly encode proteins, they can affect the expression and translation of target mRNAs through binding, thereby affecting protein production. Studies have shown that differential expression of miRNAs is closely related to various diseases including cancer. Therefore, developing a computational method to identify cancer-related miRNAs or miRNAs modules is of great significance for disease diagnosis and treatment.

[0004] Current disease-related miRNAs prediction methods are mostly knowledge-driven, such as using disease semantic similarity networks, miRNAs similarity networks, etc. However, this method relies on existing knowledge networks, and if the existing knowledge networks are not perfect or accurate, it may greatly affect the accuracy of the prediction results. SUMMARY

[0005] One or more embodiments of the present specification provide a disease-related miRNAs prediction method based on a deep autoencoder, comprising:

[0006] S1. Obtain miRNAs expression profile data of samples from the TCGA database, and perform standardization processing on the obtained miRNAs expression profile data respectively; the samples include disease samples and normal samples;

[0007] S2. Fill in the zero values in the normalized miRNAs expression profile data using a matrix decomposition method, and train a deep auto-encoder model using the filled-in miRNAs expression profile data;

[0008] S3. Train a reverse encoder based on the deep auto-encoder model, obtain the miRNA expression profile data of the disease sample and the disease correlation size according to the reverse encoder, and obtain the miRNA with high disease correlation.

[0009] One or more embodiments of the present specification provide a deep auto-encoder-based disease-related miRNAs prediction system, comprising:

[0010] The data processing module is configured to obtain miRNAs expression profile data of samples from the TCGA database, and perform standardization processing on the obtained miRNAs expression profile data respectively; the samples include disease samples and normal samples;

[0011] The model training module is configured to fill in the zero values in the normalized miRNAs expression profile data using a matrix decomposition method, and train a deep auto-encoder model using the filled-in miRNAs expression profile data;

[0012] The correlation prediction module is configured to train a reverse encoder based on the deep auto-encoder model, obtain the miRNA expression profile data of the disease sample and the disease correlation size according to the reverse encoder, and obtain the miRNA with high disease correlation.

[0013] One or more embodiments of the present specification provide an electronic device, comprising:

[0014] A processor; and

[0015] A memory arranged to store computer executable instructions that, when executed, cause the processor to implement the steps of the above-described deep auto-encoder-based disease-related miRNAs prediction method.

[0016] One or more embodiments of the present specification provide a storage medium for storing computer executable instructions that, when executed, implement the steps of the above-described deep auto-encoder-based disease-related miRNAs prediction method.

[0017] By adopting the embodiments of the present application, new disease-related miRNAs can be found, and the diagnosis and treatment of diseases can be promoted.

[0018] The above description is only a summary of the technical solutions of the present application. In order to enable one skilled in the art to better understand the technical means of the present application, the contents of the specification can be implemented, and in order to enable the above and other purposes, features and advantages of the present application to be more apparent and easy to understand, the specific embodiments of the present application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the one or more embodiments of the present specification or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present specification, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0020] Figure 1 A flowchart of a disease-related miRNAs prediction method based on a deep autoencoder provided for one or more embodiments of the present specification;

[0021] Figure 2 A composition schematic diagram of a disease-related miRNAs prediction system based on a deep autoencoder provided for one or more embodiments of the present specification;

[0022] Figure 3 A structural schematic diagram of an electronic device provided for one or more embodiments of the present specification. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the technical solutions in the one or more embodiments of the present specification, the technical solutions in the one or more embodiments of the present specification will be described clearly and completely below in combination with the drawings in the one or more embodiments of the present specification. Obviously, the described embodiments are only some embodiments of the present specification, not all embodiments. Based on the one or more embodiments of the present specification, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present document.

[0024] Method embodiments

[0025] According to the embodiments of the present application, a disease-related miRNAs prediction method based on a deep autoencoder is provided, Figure 1 A flowchart of a disease-related miRNAs prediction method based on a deep autoencoder provided for one or more embodiments of the present specification is shown as Figure 1 The disease-related miRNAs prediction method based on a deep autoencoder according to the embodiments of the present application specifically includes:

[0026] S1. Obtain miRNAs expression profile data of samples from the TCGA database, and perform standardization processing on the obtained miRNAs expression profile data respectively; the samples include disease samples and normal samples.

[0027] The disease samples include at least one disease sample, and each sample includes a plurality of miRNA expression profiles;

[0028] The standardization processing on the obtained miRNAs expression profile data is specifically:

[0029] The miRNA data of the disease samples and the normal samples are mapped to between 0 and 1, and the specific method is as follows:

[0030]

[0031] Wherein, x represents the expression value of a certain miRNA in a certain sample, x min represents the minimum expression value of miRNA in a certain sample, x max represents the maximum expression value of miRNA in a certain sample, x n represents the expression value of the standardized miRNA), and the expression value is mapped to between 0 and 1;

[0032] According to the number of disease types in the disease samples, an expression matrix of sample type x sample number is constructed together with the normal samples.

[0033] S2. Fill the zero values in the standardized miRNAs expression profile data using a matrix decomposition method, and train a deep autoencoder model using the filled miRNAs expression profile data.

[0034] The expression matrix is decomposed to obtain a first expression matrix and a second expression matrix;

[0035] The first expression matrix and the second expression matrix are multiplied to realize zero value filling in the matrix;

[0036] According to the preset learning rate and the number of training times, the above steps are cycled.

[0037] In this embodiment, the learning rate of matrix decomposition is set to 0.000002, and the number of training times is set to 5000 times.

[0038] The training of the deep autoencoder model uses the optimizer ADAM, the learning rate is 1.0x10-4, and the loss function is:

[0039]

[0040] Wherein, n represents the total number of miRNAs features, x i represents the miRNAi expression amount of miRNA representing miRNA i predicted expression amount of miRNA.

[0041] S3. Training a reverse encoder based on the deep autoencoder model, obtaining the miRNA expression profile data of the disease sample and the disease correlation size according to the reverse encoder, and obtaining the miRNA with high disease correlation.

[0042] Selecting a disease sample with a larger sample quantity in the miRNA expression profile data as a data set, each disease sample corresponding to a phenotype;

[0043] Taking the phenotype data corresponding to the disease sample in the data set as input and the miRNA expression profile data as output, training the reverse encoder.

[0044] Inputting the phenotype matrix containing each disease sample into the scoring model respectively, and each disease sample obtaining a column of scoring vector data;

[0045] Sorting all obtained scoring vector data from large to small, and the miRNAs with larger scores and ranking in the front are closely related to the disease corresponding to the input layer.

[0046] By adopting the embodiment of the present application, new disease-related miRNAs can be found, and the diagnosis and treatment of diseases can be promoted.

[0047] System embodiment

[0048] According to the embodiment of the present application, a disease-related miRNA prediction system based on a deep autoencoder is provided, Figure 2 A composition schematic diagram of a disease-related miRNA prediction system based on a deep autoencoder provided for one or more embodiments of the present application is shown in Figure 2 According to the embodiment of the present application, the disease-related miRNA prediction system based on the deep autoencoder specifically comprises:

[0049] The data processing module 20 is used for obtaining the miRNA expression profile data of the sample from the TCGA database, and performing standardization processing on the obtained miRNA expression profile data respectively; the sample includes a disease sample and a normal sample;

[0050] The model training module 22 is used for filling the zero value in the miRNA expression profile data after standardization processing by using a matrix decomposition method, and training the deep autoencoder model by using the miRNA expression profile data after filling.

[0051] The correlation prediction module 24 is configured to train a reverse encoder based on the deep auto-encoder model, obtain the miRNA expression profile data of the disease sample and the disease correlation size from the reverse encoder, and obtain the miRNA with high disease correlation.

[0052] The embodiment of the present application is a system embodiment corresponding to the above-mentioned method embodiment, and the specific operation of each module can be understood with reference to the description of the method embodiment, which will not be repeated here.

[0053] Device embodiment one

[0054] The embodiment of the present application provides an electronic device, such as Figure 3 As shown in the figure, it comprises a memory 30, a processor 32 and a computer program stored on the memory 30 and executable on the processor 32, and the computer program is executed by the processor 32 to realize the following method steps:

[0055] S1. Obtain the miRNAs expression profile data of the sample from the TCGA database, and standardize the obtained miRNAs expression profile data respectively; the sample comprises a disease sample and a normal sample;

[0056] S2. Fill the zero value in the standardized miRNAs expression profile data using a matrix decomposition method, and train a deep auto-encoder model using the filled miRNAs expression profile data;

[0057] S3. Train a reverse encoder based on the deep auto-encoder model, obtain the miRNA expression profile data of the disease sample and the disease correlation size from the reverse encoder, and obtain the miRNA with high disease correlation.

[0058] Device embodiment two

[0059] The embodiment of the present application provides a computer readable storage medium, and the computer readable storage medium stores an implementation program of information transmission, and the program is executed by the processor 32 to realize the following method steps:

[0060] S1. Obtain the miRNAs expression profile data of the sample from the TCGA database, and standardize the obtained miRNAs expression profile data respectively; the sample comprises a disease sample and a normal sample;

[0061] S2. Fill the zero value in the standardized miRNAs expression profile data using a matrix decomposition method, and train a deep auto-encoder model using the filled miRNAs expression profile data;

[0062] S3. Training a reverse encoder based on the deep autoencoder model, obtaining the miRNA expression profile data of the disease sample and the disease correlation size according to the reverse encoder, and obtaining the miRNA with high disease correlation.

[0063] The computer readable storage medium in the embodiment includes, but is not limited to, ROM, RAM, magnetic or optical disk, etc.

[0064] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting disease-related miRNAs based on deep autoencoder, characterized in that, The application relates to a method for screening miRNAs related to diseases, and a device thereof. The method comprises the following steps: S1. miRNAs expression profile data of samples are obtained from a TCGA database, and the obtained miRNAs expression profile data are standardized respectively; the samples comprise disease samples and normal samples; S2. zero values in the standardized miRNAs expression profile data are filled by using a matrix decomposition method, and a deep auto-encoder model is trained by using the filled miRNAs expression profile data; S3. a reverse encoder is trained based on the deep auto-encoder model, the correlation of miRNA expression profile data of the disease samples with diseases is obtained according to the reverse encoder, and miRNAs with high correlation with diseases are obtained. The method for training the reverse encoder based on the deep auto-encoder model comprises the following steps: Disease samples with large sample quantities in the miRNA expression profile data are selected as a data set, and each disease sample corresponds to a phenotype; phenotype data corresponding to the disease samples in the data set are taken as input, and the miRNA expression profile data are taken as output, the reverse encoder is trained, the reverse encoder is spliced with the remaining part of an original encoder, and a final scoring model is obtained; the method for obtaining the correlation of the miRNA expression profile data of the disease samples with diseases according to the reverse encoder comprises the following steps: phenotype matrices containing each disease sample are respectively input into the scoring model, and each phenotype matrix of each disease sample obtains a column of scoring vector data; 2. The method of claim 1, wherein, all obtained scoring vector data are sorted from large to small, and miRNAs with larger scores are closer to the input layer and are closely related to diseases. The disease samples comprise at least one disease sample, and each sample comprises a plurality of miRNA expression profiles; the method for standardizing the obtained miRNAs expression profile data respectively comprises the following steps: wherein, wherein x represents the expression value of a certain miRNA in a certain sample, represents the minimum expression value of a miRNA in a certain sample, represents the maximum expression value of a miRNA in a certain sample, represents the expression value of a normalized miRNA, which maps the expression value to between 0 and 1; miRNA data of the disease samples and the normal samples are mapped to 0-1, and the specific method is as follows:

3. The method of claim 2, wherein, a sample type*sample quantity expression matrix is constructed according to the number of disease types in the disease samples and the normal samples. The method for filling zero values in the standardized miRNAs expression profile data by using the matrix decomposition method comprises the following steps: the expression matrix is decomposed to obtain a first expression matrix and a second expression matrix; the first expression matrix and the second expression matrix are multiplied to realize zero value filling in the matrix; 4. The method of claim 3, wherein, the above steps are cycled according to a preset learning rate and a training number.

5. The method of claim 1, wherein, The learning rate of the matrix decomposition is 0.000002, and the training number is 5000 times. ； wherein n represents the total number of miRNAs features, represents the expression level of represents the predicted expression level of ​​ 6.A disease-associated miRNAs prediction system based on deep autoencoder, characterized in that, The deep auto-encoder model is trained by using an optimizer ADAM, the learning rate is 1.0*10-4, and the loss function is: The device comprises the following modules: a data processing module: used for obtaining miRNAs expression profile data of samples from a TCGA database, and standardizing the obtained miRNAs expression profile data respectively; the samples comprise disease samples and normal samples; The model training module is configured to fill the zero values in the normalized miRNAs expression profile data by using a matrix decomposition method, and train a deep auto-encoder model by using the filled miRNAs expression profile data. The correlation prediction module is configured to train an inverse encoder based on the deep auto-encoder model, obtain the correlation size between the miRNA expression profile data of a disease sample and the disease based on the inverse encoder, and obtain miRNAs with high correlation with the disease. The method for training the inverse encoder based on the deep auto-encoder model comprises the following steps: Selecting a disease sample with a large sample size in the miRNA expression profile data as a data set, and each disease sample corresponds to a phenotype; Taking the phenotype data corresponding to the disease sample in the data set as input and the miRNA expression profile data as output, training the inverse encoder, splicing the inverse encoder with the remaining part of the original encoder, and obtaining a final scoring model; The method for obtaining the correlation size between the miRNA expression profile data of a disease sample and the disease based on the inverse encoder comprises the following steps: Inputting a phenotype matrix containing each disease sample into the scoring model, and obtaining a column of scoring vector data for each disease sample; Sorting all obtained scoring vector data from large to small, and the miRNAs with larger scores and ranking in the front are closely related to the disease corresponding to the input layer.

7. An electronic device, comprising: The method comprises: a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to implement the steps of the deep auto-encoder based disease related miRNAs prediction method according to any one of claims 1 to 5. A computer readable medium for storing computer executable instructions that, when executed, implement the steps of the deep auto-encoder based disease related miRNAs prediction method according to any one of claims 1 to 5.

8. A storage medium, characterized by ​

Images