Diagnosis of the disease using an ant colony gene selection method based on information gain ratio using fuzzy rough sets

Abstract

With the advancement of metagenome data mining science has become focused on microarrays. Microarrays are datasets with a large number of genes that are usually irrelevant to the output class; hence, the process of gene selection or feature selection is essential. So, it follows that you can remove redundant genes and increase the speed and accuracy of classification. After applying the gene selection, the dataset is reduced and detection of differentially abundant genes facilitated with more accuracy. This will, in turn, increases the power of genes which are correctly detected statistically differentially abundant in two or more phenotypes. The method presented in this study is a two-stage method for functional analysis of metagenomes. The first stage uses a combination of the filter and wrapper gene selection method, which includes the ant colony algorithm and utilizes fuzzy rough sets to calculate the information gain ratio as an evaluation measure in the ant colony algorithm. The set of features from the first stage is used as input in the second stage, and then the negative binomial distribution is used to detect genes which are statistically differentially abundant in two or more phenotypes. Applying the proposed method on a microarray dataset it becomes clear that the proposed method increases the accuracy of the classifier and selects a subset of genes that have a minimum length and maximum accuracy.

Keywords

• microarray
• gene selection
• phenotype
• fuzzy rough set
• ant colony optimization
• Gain ratio