CRIA: An Interactive Gene Selection Algorithm for Cancers Prediction Based on Copy Number Variations

Abstract

Genomic copy number variations (CNVs) are among the most important structural variations of genes found to be related to the risk of individual cancer and therefore they can be utilized to provide a clue to the research on the formation and progression of cancer. In this paper, an improved computational gene selection algorithm called CRIA (correlation-redundancy and interaction analysis based on gene selection algorithm) is introduced to screen genes that are closely related to cancer from the whole genome based on the value of gene CNVs. The CRIA algorithm mainly consists of two parts. Firstly, the main effect feature is selected out from the original feature set that has the largest correlation with the class label. Secondly, after the analysis involving correlation, redundancy and interaction for each feature in the candidate feature set, we choose the feature that maximizes the value of the custom selection criterion and add it into the selected feature set and then remove it from the candidate feature set in each selection round. Based on the real datasets, CRIA selects the top 200 genes to predict the type of cancer. The experiments' results of our research show that, compared with the state-of-the-art related methods, the CRIA algorithm can extract the key features of CNVs and a better classification performance can be achieved based on them. In addition, the interpretable genes highly related to cancer can be known, which may provide new clues at the genetic level for the treatment of the cancer.

Radiogenomics Reveals Correlation between Quantitative Texture Radiomic Features of Biparametric MRI and Hypoxia-Related Gene Expression in Men with Localised Prostate Cancer

“…A cross-sectional gene-oriented query on cBioprtal ( , accessed on 9 June 2021) (34) was conducted to validate the overexpressed hypoxic gene based on the database. cBioportal, a widely used web interface that gives access to public cancer genomics datasets, was utilised to retrieve the Cancer Genome Atlas data [ 34 , 35 , 36 ]. The focus gene-query for prostate cancer was chosen, with genomic information starting with HUGO genes.…”