Type2 soft biclustering framework for Alzheimer microarray

Abstract

Microarray technology is a powerful tool that enables simultaneous analysis of the expression level of a large number of genes for different samples. Reliable information on gene expression level is much needed in the health system as it is widely used to predict, diagnose, and treat human diseases (e.g., Alzheimer's). For the analysis of the microarray dataset, biclustering is known to be a highly capable approach, however some characteristics of the dataset including high dimensionality, noise, uncertainty, and complex biological processes need to be handled properly. Concerning these characteristics, the current paper proposes a novel two-stage biclustering framework based on soft clustering and a metaheuristic technique. The integration of the two stages ensures a reliable search process to find similar expression patterns concerning gene expression characteristics. The proposed framework employs fuzzy and possibilistic clustering along with Type2-Fuzzy Sets theory to handle high-level uncertainty, noise, and outliers in microarray datasets. Considering the NP-hard nature of the biclustering method, the proposed framework incorporates the Genetic Algorithm with a unique chromosome representation, fitness function, and modification mechanisms. Real microarray datasets of Alzheimer's Disease have been used to evaluate the proposed framework. The comparative analysis of different versions of our proposed framework and some well-known biclustering methods demonstrates that the proposed framework is superior in terms of some indices including the mean squared residual and variance indices. The final results are further evaluated using the defined fitness function, which indicates the better performance of our possibilistic-based biclustering methods.