An Optimization on Bicluster Algorithm for Gene Expression Data

Biclustering enables the identification of gene groups that cannot be discovered using classical clustering approaches, which typically operate on all experimental conditions simultaneously. A bicluster is a combination of a set of genes and a group of samples, where the genes exhibit similarity within the samples, and vice versa. When a sample shows the activation in multiple pathways, it can belong to different biclusters. The biclustering method processes two types of data matrices: binary and non-binary. Biclustering algorithms that handle binary data often struggle to strike a satisfactory equilibrium between execution speed and efficiency. This paper presents a novel biclustering algorithm for binary matrix for generating, assessing and validating bicluster, by incorporating a straightforward binary reference model for comparison. The proposed methodology is applied to gene expression data, by utilizing the neighbour joining difference matrix to identify similar genes. Adjacency matrix enables the clustering of genes with similar reactions under various conditions into coherent clusters, which plays a crucial influence in the subsequent analysis of genes. This approach exhibits improved time complexity and relevance scores compared to other biclust algorithms like Bibit and Qubic.