Evaluation of the Gene Expression Profiles Complex Proximity Metric Effectiveness Based on a Hybrid Technique of Gene Expression Data Extraction

Abstract

Gene expression data processing in order to develop the systems of complex diseases diagnostic or/and gene regulatory networks (GRN) reconstruction is one of the actual direction of modern bioinformatics. One of the important stages of this problem solving is an extraction of mutually correlated gene expression profiles (GEP) considering the used proximity metric. Within the framework of our research, we evaluate the complex metric of GEP proximity calculated as the combination of modified mutual information criterion and Pearson's chi-squared test using OPTICS clustering algorithm implemented using principles of the objective clustering inductive technique (OCIT). The examined objects classification accuracy was used as the main criterion to access the applied method effectiveness. The simulation results have shown that the proposed technique allows us to form an optimal GEP cluster structure in terms of maximum values of the patterns classification accuracy quality criterion.