Efficient query-driven biclustering of gene expression data using Probabilistic Relational Models

Biclustering is an increasingly popular technique to identify gene regulatory modules that are linked to biological processes. We describe a novel method, called ProBic, that was developed within the framework of Probabilistic Relational Models (PRMs). ProBic is an efficient biclustering algorithm that simultaneously identifies a set of potentially overlapping biclusters in a gene expression dataset and which can be used both in a query-driven and a global setting. The model naturally deals with missing values. Robust sets of biclusters are obtained due to the explicit modeling of noise. The maximum likelihood solution is approximated using an Expectation-Maximization strategy. ProBic was applied to various synthetic gene expression datasets and the results for synthetic data confirmed that ProBic can successfully identify biclusters under various noise levels, overlap and missing values in both the query-driven and global setting. Additional expert knowledge can be introduced through a number of prior distribution parameters. Default settings were shown to be applicable for a wide range of different datasets. Our results on synthetic data show that PRMs can be used to identify overlapping biclusters in an efficient and robust manner. keywords: biclustering, probabilistic relational model, gene expression, regulatory module, expectation-maximization. ∗corresponding author