Gene selection and cancer classification methods based on Monte Carlo and nonnegative matrix factorization

Abstract

A gene selection and cancer classification method based on Monte Carlo and non-negative matrix factorization, using Monte Carlo method to generate multiple gene subsets from the gene expression data of the original sample; each subset is passed through the non-negative matrix factorization method It is decomposed into a coefficient matrix and a base matrix; in each non-negative matrix factorization iteration, if the sparsity of a certain sample in the base matrix is ​​less than the minimum sparsity of the original sample, the elements in the sample are gradually replaced by zeros from small to large, until the The sparsity is no longer less than the minimum sparsity of the original sample; judge convergence; after iterative convergence, the gene score is used to judge the importance of the gene: arrange the genes in descending order according to the score, select the genes in the sequence one by one to build a series of models, and each model passes Ten-fold cross-validation method correction; the model with the best accuracy is used for prediction. The method can effectively identify biomarkers in genes, and the model established by the identified biomarkers can be used to effectively predict phenotypes in new cancers.

Description

technical field [0001] The invention belongs to the technical field of chemometrics, and relates to a gene selection and cancer classification method based on Monte Carlo and non-negative matrix factorization. Background technique [0002] Cancer classification is a key issue in clinical research to identify biomarkers and cure malignancies. Gene expression profiles obtained by microarray gene chip technology have been successfully applied to identify biomarkers and classify cancer samples. [0003] Gene expression profiles reflect biological information through a large amount of genetic data. All genetic data in gene expression profiles can be considered as potential biomarkers. Some important biomarkers in genetic data and features of genetic data can be used to accurately predict the phenotype of new tumors. However, if all the data is used, classification will run into problems with high-dimensional data. [0004] Nonnegative matrix factorization (NMF) can generate n...

AI Technical Summary

Problems solved by technology

However, the sparse basis vectors of the genetic data array obtained through sparseness will lose a lot of important information of the original genetic data, and the greater the sparsity, the more information is lost

Images