Development of Methodology for Trait Specific Genes Identification

Abstract

Comprehensive profiling of biological system is being continuously done using expression data obtained through high-throughput technologies, such as gene expression (GE) data, protein expression data and medical imaging data. The resulting data sets obtained through these technologies are huge in size and have several common characteristics making their analysis challenging. Many times the number of genes is much larger than the number of sample and the relevant informative genes which are associated with the outcome are less in the data sets. It is important to select most relevant genes related to condition class from thousands of genes with the help of appropriate statistical and computational techniques. These high dimensional data can be transformed into a meaningful representation of reduced dimensions (informative data) using dimensionality reduction techniques. Informative gene selection plays a bigger role in removing irrelevant and redundant genes from the data set. In this study methodology for obtaining relevant set of trait specific genes from gene expression data by applying combination of two conventional machine learning algorithms, support vector machine (SVM) and a genetic algorithm (GA) has been developed. Using SVM as the classifier performance and the Genetic algorithm for feature selection, a set of informative genes set can be obtained. The classification accuracy of the obtained genes set from the developed methodology was compared with the genes set obtained from methods such as Boot-MRMR, MRMR, t-score and F-score of R- package “GSAQ”. It has been observed that the performance of the developed methodology is better as compared to above given techniques for selecting robust set of informative genes. Based on this proposed approach, an R package, i.e., TSGS (https://cran.r-project.org/package=TSGS) has also been developed.