A hybrid stellar mass black-hole optimization framework for finding significant biclusters using average Kendall rank correlation

Microarray gene expression data are high-dimensional and complex, with patterns that may appear only under specific conditions. Traditional clustering often misses these local patterns, whereas biclustering can reveal groups of genes with coordinated expression across particular conditions. In this paper, we propose a biclustering approach using average Kendall correlation, which captures nonlinear and monotonic relationships often overlooked by standard measures like Euclidean distance or Pearson correlation. To efficiently search for optimal biclusters, we implement a modified stellar mass black-hole optimization (MSBO) approach that integrates the Nelder-Mead simplex method with Lévy flight to enhance both local and global search capabilities. The proposed technique is validated on couple of widely used benchmark gene expression datasets, namely the yeast cell cycle and lymphoma. The biological importance of the identified biclusters is evaluated with the help of the gene ontology (GO) database. Experimental results demonstrate that our method outperforms traditional approaches in identifying statistically significant and biologically relevant biclusters, achieving a p-value of 3.73 × 10^-16. These findings address the pressing need for more effective biclustering techniques in the study of microarray data.