Research On Anti-cancer Drug Sensitivity Prediction Method Integrating Multi-omics Data

In the context of the development of precision medicine technology,it is crucial to study the response of cancer patients to treatment based on their own genomic information and clinical phenotypic characteristics.The key to achieve this goal is to accurately predict the sensitivity of anticancer drugs to patients.Since the publication of large-scale drug screening data,many methods for predicting the sensitivity of anticancer drugs have been proposed,but most of them are based on the characteristic of gene expression in cells.However,we know that in addition to gene expression,information describing the genomic state of a sample,such as mutations,copy number variations,DNA methylation,etc.also provide crucial information for determining drug sensitivity.In this context,it is of great scientific significance to establish a prediction model of anticancer drug sensitivity by integrating multi-group data.In this paper,two network models for drug sensitivity prediction were established by integrating gene expression,copy number variation,methylation and small RNA expression of cell lines.The first model is GKEDM(Gaussian kernel Euclidean distance model)multi-group network model.The model firstly calculated the correlation between cells,and predicted the sensitivity of drugs to cell lines by multi-group characteristics such as gene expression,copy number and small RNA expression.Then the mesh method is used to optimize the weight of multi-group features.Finally,the network model is built by feature integration.The second model is GKCCM(Gaussian kernel correlation coefficient model)multi-group network model.This model to establish the overall train of thought and the first model is basically the same,the main difference is that the model using the Gaussian kernel function to calculate Pearson correlation coefficient correlation between cells,finally calculated the two models of drug sensitivity to cell lines and real and estimated values of correlation coefficient,and its results with only compares the forecasting model based on gene expression.The overall structure of this paper is as follows: the first chapter mainly introduces the significance of drug sensitivity,the research status and the overall framework of this paper.The second chapter mainly introduces the source of data and the process of model construction.The data source includes the download and preprocessing of the data.In terms of model construction,the basic model structure,parameter selection method,model evaluation method and specific model research are mainly introduced.The third chapter analyzes the results of the model from two aspects.Firstly,the results of single factor analysis were compared with the results of integrated multi-group model.In addition,we supplemented the missing data of drug sensitivity in CCLE data to further verify the effectiveness of the model.In the fourth chapter,the model of this paper is summarized and prospected,and the advantages and disadvantages of the anticancer drug sensitivity prediction network model integrating multi-group data and the future research direction are discussed.