CN-121999868-A - Drug resistance key gene screening method, device, electronic equipment and storage medium

CN121999868ACN 121999868 ACN121999868 ACN 121999868ACN-121999868-A

Abstract

The invention relates to the technical field of gene screening, in particular to a drug resistance key gene screening method, a device, electronic equipment and a storage medium, which comprise the following steps: constructing a drug resistance gene extraction task for obtaining a drug resistance gene expression profile in an mRNA expression profile by utilizing a first neural network, constructing a drug resistance key gene extraction task for obtaining a drug resistance key gene expression profile in the drug resistance gene expression profile by utilizing a second neural network, and carrying out joint learning on the drug resistance gene extraction task and the drug resistance key gene extraction task by a multi-task learning mode to construct a drug resistance key gene screening task for obtaining the drug resistance key gene expression profile in the mRNA expression profile. The invention can improve the mastering accuracy of the gene expression level, thereby improving the accuracy of gene screening and improving the accuracy of the finally obtained drug resistance key genes.

Inventors

HE YANG
CHEN CHUANRONG
HE XIAOYING
HU QIN

Assignees

皖南医学院第一附属医院（皖南医学院弋矶山医院）

Dates

Publication Date: 20260508
Application Date: 20240202

Claims (10)

1. The drug resistance key gene screening method is characterized by comprising the following steps: constructing an in vitro breast cancer drug-resistant cell line 4T1/PTX of triple negative breast cancer TNBC; detecting mRNA expression profiles of the in vitro breast cancer drug resistant cell line 4T1/PTX using high throughput sequencing techniques; Performing characteristic extraction on an mRNA expression profile of an in-vitro breast cancer drug-resistant cell line 4T1/PTX by using a first neural network, and constructing a drug-resistant gene extraction task for obtaining a drug-resistant gene expression profile in the mRNA expression profile; performing feature extraction on the drug resistance gene expression profile by using a second neural network to construct a drug resistance key gene extraction task for obtaining a drug resistance key gene expression profile in the drug resistance gene expression profile; And performing joint learning on the drug resistance gene extraction task and the drug resistance key gene extraction task in a multi-task learning mode to construct a drug resistance key gene screening task for obtaining a drug resistance key gene expression profile in the mRNA expression profile.
2. The method for screening drug-resistant key genes according to claim 1, wherein the method for constructing the in vitro breast cancer drug-resistant cell line 4T1/PTX of the triple negative breast cancer TNBC comprises the following steps: first, cells were cultured at 1/10 drug concentration of IC50 for two weeks; Then, the drug concentration was increased to 1/8 of the IC50, and the cells were cultured for two weeks; finally, continuously increasing the drug concentration, and culturing for 4-6 months to induce an in vitro breast cancer drug-resistant cell line 4T1/PTX.
3. The method for screening drug resistance key genes according to claim 1, wherein the construction method for the drug resistance gene extraction task comprises the following steps: obtaining mRNA expression profiles of n breast cancer drug-resistant cells in an in vitro breast cancer drug-resistant cell line 4T 1/PTX; constructing a first neural network formed by juxtaposing n CNN neural network structures; Respectively taking mRNA expression profiles of n breast cancer drug-resistant cells as input items of n CNN neural network structures in a first neural network, respectively correspondingly outputting the mRNA expression profiles of the n breast cancer drug-resistant cells by the n CNN neural networks in the first neural network, and taking the mRNA expression profiles of the n breast cancer drug-resistant cells as the drug-resistant gene expression profiles of the n breast cancer drug-resistant cells to obtain the drug-resistant gene extraction task; the drug resistance gene extraction task is as follows: Wherein G1 is the drug resistance gene expression profile of the 1 st breast cancer drug resistant cell, G2 is the drug resistance gene expression profile of the 2 nd breast cancer drug resistant cell, gn is the drug resistance gene expression profile of the n th breast cancer drug resistant cell, G1 is the mRNA expression profile of the 1 st breast cancer drug resistant cell, G2 is the mRNA expression profile of the 2 nd breast cancer drug resistant cell, gn is the mRNA expression profile of the n st breast cancer drug resistant cell, CNN1 is the 1 st CNN neural network structure in the first neural network, CNN2 is the 2 nd CNN neural network structure in the first neural network, CNNn is the n th CNN neural network structure in the first neural network; Taking a loss function of the first neural network as a task target of the drug resistance gene extraction task, wherein the task target of the drug resistance gene extraction task is as follows: lossg=MSE(g1,g2,...gn); Wherein lossg is a task target of a drug resistance gene extraction task, MSE is a mean square error operation formula, MSE (g 1, g2,..gn) is a mean square error between g1, g2,..gn.
4. The method for screening drug resistance key genes according to claim 3, wherein the construction method for the drug resistance key gene extraction task comprises the following steps: Constructing a second neural network formed by paralleling n twin neural network structures; Respectively taking mRNA expression profiles of n breast cancer drug-resistant cells as input items of one neural network structure in n twin neural network structures in a second neural network, respectively correspondingly outputting the mRNA expression profiles of the n breast cancer drug-resistant cells by the one neural network structure in the n twin neural network structures in the second neural network, and taking the mRNA expression profiles of the n breast cancer drug-resistant cells as drug-resistant gene expression profiles of the n breast cancer drug-resistant cells; The drug resistance gene expression spectrums of n breast cancer drug resistant cells are respectively used as input items of another neural network structure in n twin neural network structures in a second neural network, the other neural network structure in the n twin neural network structures in the second neural network is respectively correspondingly output, and the spectrum characteristics of the drug resistance key genes are represented in the drug resistance gene expression spectrums of n breast cancer drug resistant cells and are used as drug resistance key gene expression spectrums of n breast cancer drug resistant cells; obtaining a drug resistance key gene extraction task; The drug resistance key gene extraction task is as follows: wherein model1, model2 and modeln are respectively 1 st, 2 nd and n th twin neural network structures in the second neural network, G1, G2 and Gn are respectively drug resistance gene expression profiles of 1 st, 2 nd and n th breast cancer drug resistant cells, s1, s2 and sn are drug resistance key gene expression profiles of 1 st, 2 nd and n th breast cancer drug resistant cells, G1, G2 and Gn are mRNA expression profiles of 1 st, 2 nd and n th breast cancer drug resistant cells, CNN11, CNN21 and CNNn1 are one CNN neural network structure of 1 st, 2 nd twin neural network structures in the second neural network, CNN12, CNN22 and CNNn2 are another CNN neural network structure of 1 st, 2 nd twin neural network structures in the second neural network; Taking a loss function of a second neural network as a task target of the drug resistance key gene extraction task, wherein the task target of the drug resistance key gene extraction task is as follows: wherein lossgs is a task target of a drug resistance key gene extraction task, MSE is a mean square error operation formula, MSE (gi, si) is a mean square error between gi and si, and i is a counting variable.
5. The method for screening drug resistance genes according to claim 4, wherein the construction method for the drug resistance gene screening task comprises the steps of: sharing the bottom network layers of the drug resistance gene extraction task and the drug resistance key gene extraction task by utilizing a multi-task learning mechanism; And co-building a task target of a drug resistance gene extraction task and a task target of a drug resistance key gene extraction task to obtain a task target of a drug resistance key gene screening task.
6. The method for screening drug resistance genes according to claim 5, wherein the task objective of the drug resistance gene screening task is: Loss(t)=wg(t)*lossg(t)+wgs(t)*lossgs(t); in the formula, loss (t) is a task target of a drug resistance key gene screening task in a t-th task stage, loss (t) is a task target of a drug resistance gene extraction task in the t-th task stage, loss (t) is a task target of a drug resistance key gene extraction task in the t-th task stage, wg (t) is a task weight of a drug resistance gene extraction task in the t-th task stage, wgs (t) is a task weight of a drug resistance key gene extraction task in the t-th task stage, and t is a counting variable; wg(t)=-[1-Ke(t)]log[Ke(t)]; wgs(t)=-[1-Kr(t)]log[Kr(t)]; Wherein Ke (t) is the accuracy of the drug resistance gene extraction task in the t-th task stage, and Kr (t) is the accuracy of the drug resistance key gene extraction task in the t-th task stage.
7. The method for screening drug-resistant key genes according to claim 2, wherein mRNA expression profiles of n breast cancer drug-resistant cells are normalized.
8. A drug resistance key gene screening device, comprising: A biological acquisition module for constructing an in vitro breast cancer drug-resistant cell line 4T1/PTX of triple negative breast cancer TNBC, and For detecting mRNA expression profiles of the in vitro breast cancer drug resistant cell line 4T1/PTX using high throughput sequencing techniques; the task building module is used for carrying out characteristic extraction on an mRNA expression profile of an in-vitro breast cancer drug-resistant cell line 4T1/PTX by using a first neural network, and constructing a drug resistance gene extraction task for obtaining a drug resistance gene expression profile in the mRNA expression profile; The characteristic extraction is performed on the drug resistance gene expression profile by using a second neural network to construct a drug resistance key gene extraction task for obtaining the drug resistance key gene expression profile in the drug resistance gene expression profile, and The drug resistance gene screening task is used for carrying out joint learning on a drug resistance gene extraction task and a drug resistance key gene extraction task in a multi-task learning mode to construct a drug resistance key gene screening task for obtaining a drug resistance key gene expression profile in an mRNA expression profile; and the gene screening module is used for utilizing a drug resistance key gene screening task to obtain a drug resistance key gene expression profile.
9. An electronic device comprising a memory and a processor, The memory being coupled to the processor, the memory being for storing a program, the processor invoking the program stored in the memory to perform the method of any of claims 1-7.
10. A storage medium having stored thereon a computer program which, when executed by a computer, performs the method of any of claims 1-7.

Description

Drug resistance key gene screening method, device, electronic equipment and storage medium Technical Field The invention relates to the technical field of gene screening, in particular to a drug resistance key gene screening method, a device, electronic equipment and a storage medium. Background Drug resistance of tumor cells to therapeutic drugs is an unavoidable event in clinical tumor treatment, and limits the effect of drug treatment, thereby affecting the cure of cancer, so that it is necessary to screen out drug resistance key genes that have a key effect on drug resistance of tumor cells for research. In the existing methods for screening drug resistance key genes based on transcriptome data, a gene differential expression analysis method is generally adopted. However, the gene differential expression analysis method only considers the difference of the expression level of the genes, namely, when the drug resistance key genes are screened, the accuracy is inaccurate due to the fact that the expression level of the genes is mastered, so that the accuracy of the gene screening is affected, and the accuracy of the finally obtained drug resistance key genes is low. Disclosure of Invention The invention aims to provide a drug resistance key gene screening method, a device, electronic equipment and a storage medium, which are used for solving the technical problems that the accuracy of gene screening is affected due to inaccurate mastering precision of gene expression level in the prior art, and the accuracy of the finally obtained drug resistance key gene is low. In order to solve the technical problems, the invention specifically provides the following technical scheme: in a first aspect, a method for screening a drug resistance key gene comprises the steps of: constructing an in vitro breast cancer drug-resistant cell line 4T1/PTX of triple negative breast cancer TNBC; detecting mRNA expression profiles of the in vitro breast cancer drug resistant cell line 4T1/PTX using high throughput sequencing techniques; Performing characteristic extraction on an mRNA expression profile of an in-vitro breast cancer drug-resistant cell line 4T1/PTX by using a first neural network, and constructing a drug-resistant gene extraction task for obtaining a drug-resistant gene expression profile in the mRNA expression profile; performing feature extraction on the drug resistance gene expression profile by using a second neural network to construct a drug resistance key gene extraction task for obtaining a drug resistance key gene expression profile in the drug resistance gene expression profile; And performing joint learning on the drug resistance gene extraction task and the drug resistance key gene extraction task in a multi-task learning mode to construct a drug resistance key gene screening task for obtaining a drug resistance key gene expression profile in the mRNA expression profile. As a preferred scheme of the invention, the construction method of the in vitro breast cancer drug-resistant cell line 4T1/PTX of the triple negative breast cancer TNBC comprises the following steps: first, cells were cultured at 1/10 drug concentration of IC50 for two weeks; Then, the drug concentration was increased to 1/8 of the IC50, and the cells were cultured for two weeks; finally, continuously increasing the drug concentration, and culturing for 4-6 months to induce an in vitro breast cancer drug-resistant cell line 4T1/PTX. As a preferable scheme of the invention, the construction method of the drug resistance gene extraction task comprises the following steps: obtaining mRNA expression profiles of n breast cancer drug-resistant cells in an in vitro breast cancer drug-resistant cell line 4T 1/PTX; constructing a first neural network formed by juxtaposing n CNN neural network structures; Respectively taking mRNA expression profiles of n breast cancer drug-resistant cells as input items of n CNN neural network structures in a first neural network, respectively correspondingly outputting the mRNA expression profiles of the n breast cancer drug-resistant cells by the n CNN neural networks in the first neural network, and taking the mRNA expression profiles of the n breast cancer drug-resistant cells as the drug-resistant gene expression profiles of the n breast cancer drug-resistant cells to obtain the drug-resistant gene extraction task; the drug resistance gene extraction task is as follows: Wherein G1 is the drug resistance gene expression profile of the 1 st breast cancer drug resistant cell, G2 is the drug resistance gene expression profile of the 2 nd breast cancer drug resistant cell, gn is the drug resistance gene expression profile of the n th breast cancer drug resistant cell, G1 is the mRNA expression profile of the 1 st breast cancer drug resistant cell, G2 is the mRNA expression profile of the 2 nd breast cancer drug resistant cell, gn is the mRNA expression profile of the n st breast cancer drug resistant cell, CNN1 is