Gene Selection for Single-Cell RNA-Seq Data Based on Information Gain and Genetic Algorithm

Abstract

Single-cell RNA-seq data often contain tens of and thousands genes, while too many of them are redundant genes as well as inferior genes. In this study, the information gain is used to coarsely remove those redundant and inferior genes, then a new designed genetic algorithm with a dynamic crossover operator is used to finely select the most important genes. The new feature selection algorithm is abbreviated as IGGA. The difference between the IGGA and the existing methods lies in that IGGA is designed at the first time to select genes from single-cell RNA-seq data. Experimental results performing on several real datasets demonstrate that the proposed algorithm can efficiently select the most important genes.