bioRxiv2021
Comprehensive molecular characterization of cancer subtypes is essential for predicting clinical outcomes and searching for personalized treatments. We present bnClustOmics, a statistical model and computational tool for multi-omics unsupervised clustering, which serves a dual purpose: Clustering patient samples based on a Bayesian network mixture model and learning the networks of omics variables representing these clusters. The discovered networks encode interactions among all omics variables and provide a molecular characterization of each patient subgroup. We conducted simulation studies that demonstrated the advantages of our approach compared to other clustering methods in the case where the generative model is a mixture of Bayesian networks. We applied bnClustOmics to a hepatocellular carcinoma (HCC) dataset comprising genome (mutation and copy number), transcriptome, proteome, and phosphoproteome data. We identified three main HCC subtypes together with molecular characteristics, some of which are associated with survival even when adjusting for the clinical stage. Cluster-specific networks shed light on the links between genotypes and molecular phenotypes of samples within their respective clusters and suggest targets for personalized treatments. Author summaryMulti-omics approaches to cancer subtyping can provide more insights into molecular changes in tumors compared to single-omics approaches. However, most multi-omics clustering methods do not take into account that gene products interact, for example, as parts of protein complexes or signaling networks. Here we present bnClustOmics, a Bayesian network mixture model for unsupervised clustering of multi-omics data, which can represent dependencies among molecular changes of various omics types explicitly. Unlike other approaches that use data from public interaction databases as ground truth, bnClustOmics learns the dependencies between genes from the analyzed multi-omics dataset. At the same time, our approach can also account for prior knowledge from public interaction databases and use it to guide network learning without losing the ability to learn new dependencies. We applied bnClustOmics to a multi-omics HCC dataset and identified three subtypes similar to those identified in other HCC studies. The cluster-specific networks learned by bnClustOmics revealed additional insights into the molecular characterization of the discovered subgroups and highlighted the changes in signaling networks leading to distinct HCC phenotypes.