An Integrative Learning-Based Biclustering Algorithm for Cancer Multi-Omics Data

Integrative analysis of cancer multi-omics data can provide a more comprehensive understanding of the underlying biomolecular mechanisms of cancer. To this end, a novel integrative learning of biclustering algorithm for cancer omics data is proposed, enabling integrative analysis of miRNA and gene expression data. The proposed algorithm first employs a multiple-output sparse group Lasso regression model (MSGL) to integrate miRNA and gene expression data, constructing a sparse miRNA-gene regulatory matrix. Subsequently, this regulatory matrix is incorporated into a biclustering algorithm with <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$l_{1 / 2}$</tex>-norm sparse graph regularization to identify biologically meaningful miRNA-gene regulatory modules. Experimental results on different cancer multi-omics datasets demonstrate that the miRNA-gene regulatory modules obtained by the proposed algorithm are more likely to exhibit co-regulatory relationships. The proposed algorithm provides a novel perspective for cancer pathology analysis, facilitating a deeper understanding of the complex coregulatory interactions between genes and miRNAs.