Identification and analysis of gene clusters in biological data

Abstract

In this work, a methodology for constructing and analysing gene clusters based on gene expression data is proposed. This approach uses data mining algorithms and machine learning methods to discover relationships between genes associated to biological knowledge and regulatory activities for expression experiments conducted under particular conditions of interest. Such gene groups were constructed based on a similarity matrix defined from nonlinear relationships given by a constructed kernel on each data type. Once groups were formed, representative transcription factor binding sites were searched in each cluster in order to categorize them and to find patterns and relationships between genes and transcription factor binding sites. Accordingly, interesting transcription factor binding sites related to gene expression experimental conditions were detected. This confirmed common gene regulation of known binding sites.