Bayesian Network-Based Clustering of Multi-Omics Data
Description
Unsupervised Bayesian network-based clustering of multi-omics data. Both binary and continuous data types
are allowed as inputs. The package serves a dual purpose: it clusters (patient) samples and learns the multi-omics networks that characterize discovered
clusters. Prior network knowledge (e.g., public interaction databases) can be included via blacklisting and
penalization matrices. For clustering, the EM algorithm is employed. For structure search at the M-step,
the Bayesian approach is used. The output includes membership assignments of samples, cluster-specific MAP networks, and posterior probabilities
of all edges in the discovered networks. In addition to likelihood, AIC and BIC scores are returned. They can be used for choosing the number
of clusters.
References:
P. Suter et al. (2021) ,
J. Kuipers and P. Suter and G. Moffa (2022) ,
J. Kuipers et al. (2018) .