The Algorithms And Models For Cancer Driver Mutation Identification

DN A mutations could lead to the occurrence and progression of cancers,however,some types of mutations are not exactly associated with cancer development.Mutations contributing to the growth of cancer cells and positively selected in cancer microenvironment are defined as driver mutations.On the contrary,mutations that can not result in cancer development are categorized into passenger mutations.The goal of cancer genome analysis is identifying cancer genes with driver mutations,hence differentiating driver and passenger mutations is an important research topic.Traditional methods often used the structural characteristics of mutation to screen driver mutations.With the advance of genome sequencing techniques,previous approaches were difficult to analyze high-throughput sequencing data,which limited the power for driver mutation identification.With the development of bioinformatics technologies,identifying driver mutations during cancer evolution via computational methods such as machine learning is of great significance.This study aims at constructing bioinformatics models for cancer driver mutation identification.In the first step,we selected the features associated with mutations.Then we proposed the cancer-specific computational model.Based on the extraction of mutation features for cancers,we used the MRMR algorithm to remove redundancy and got the optimal feature subset.Finally,we built the support vector machine-based model Driver-Ⅰ,and applied it to identify driver mutations for prostate cancer.The result of performance comparison indicated the good accuracy and generality of the model.This study identified driver mutations of cancers based on machine learning methods,which is propitious to the understanding of molecular mechanisms related to cancer development,and it may contribute to the discovery of novel biomarkers and drug targets for cancers.