Feature selection in bioinformatics

Abstract

In bioinformatics, there are often a large number of input features. For example, there are millions of single nucleotide polymorphisms (SNPs) that are genetic variations which determine the difference between any two unrelated individuals. In microarrays, thousands of genes can be profiled in each test. It is important to find out which input features (e.g., SNPs or genes) are useful in classification of a certain group of people or diagnosis of a given disease. In this paper, we investigate some powerful feature selection techniques and apply them to problems in bioinformatics. We are able to identify a very small number of input features sufficient for tasks at hand and we demonstrate this with some real-world data.