Formation of Subsets of Co-expressed Gene Expression Profiles Based on Joint Use of Fuzzy Inference System, Statistical Criteria and Shannon Entropy

Abstract

AbstractThe paper presents the results of the research regarding the application of a fuzzy logic inference system to form the co-expressed gene expression profiles based on the joint use of Shannon entropy and statistical criteria . The allocation of co-expressed genes can allow us to increase the disease diagnosis accuracy on the one hand and, reconstruct the qualitative gene regulatory networks on the other hand. To solve this problem, we have proposed the joint use of the fuzzy logic inference system and random forest classifier. In beginning, we have calculated for each of the gene expression profiles the maximum expression values, variance and Shannon entropy. These parameters were used as the input ones for the fuzzy logic inference system. After setting the fuzzy membership functions for both the input and output parameters, the model formalization including fuzzy rules formation, we have applied the model to gene expression data which included initially the 54675 genes for 156 patients examined at the early stage of lung cancer. As a result of this step implementation, we have obtained the four subsets of gene expression profiles with various significance levels. To confirm the obtained results, we have applied the classification procedure to investigated samples that included as the attributes the allocated genes. The analysis of the classification quality criteria allows us to conclude about the high effectiveness of the proposed technique to solve this type of task.KeywordsGene expression profilesFuzzy modelStatistical criteriaShannon entropyCo-expressed genesClassification quality criteriaROC analysis