Entropy-based fuzzy hybrid framework for gene prediction network – an application to identify and rank the biomarkers for human lung adenocarcinoma

Abstract

ABSTRACTCancer occurs because of an anomaly in cell structure due to the uncontrolled behavior of genes. The approach to reduce and identify the set of dominant biomarkers for malfunctioning genes which may lead to various types of cancer has been identified. A reduced Gene Prediction Graph depicting the dominant genes along with the extent of their interdependence is plotted both for a healthy person and a diseased individual. The genes are ranked with decreasing order of relevance. A detailed comparative study between the graphs detects the set of malfunctioning genes with decreasing order of influence. The theory of Information Gain, Fuzzy Logic, Graph theory-based Algorithm, Entropy, and Symmetric Uncertainty has been applied in the development of the proposed model named as Entropy-based fuzzy hybrid framework for gene prediction network. The entire algorithm is demonstrated in fuzzy framework of gene expression datasets. A detailed comparative study has been done with some of the existing methodologies, viz., Support Vector Machine approach based on Recursive Feature Elimination, Hidden Markov Model, Gaussian Mixture Model, Fuzzy Composite Association Rule Mining, Frequent Pattern-Growth, Apriori. The performance of the methodology has been demonstrated on the lung adenocarcinoma datasets. The results are appropriately validated using top 100 genes of National Center for Biotechnology Information database, for finding out the true positives.