A Penalized Regression-Based Biclustering Approach in Gene Expression Data

Clustering serves as a pivotal instrument in the realm of gene expression data analysis. This paper proposes a Biclustering Coefficient Estimation (BCE) method to identify groups in the individuals and genes. An alternating direction method of multipliers (ADMM) algorithm with a double fusion penalty is developed to solve the problem. The authors rigorously establish the oracle properties for the proposed penalized estimator. Numerical studies, including simulations and analysis of a lung adenocarcinoma dataset, suggest that the proposed method is expected to simultaneously recover reasonable potential groups of samples and covariates and provide satisfactory estimates of group coefficients.