Generalized Bayesian Factor Analysis for Integrative Clustering with Application to Multi-Modal Omics Data

• Activity Number: 485 - Bayesian Latent Variable Methods for Life Sciences
• Type: Contributed
• Date/Time: Wednesday, August 1, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Eun Jeong Min* and Changgee Chang and Qi Long
• Companies: University of Pennsylvania and University of Pennsylvania and University of Pennsylvania
• Keywords: Integrative Clustering; Multi-modal omics; Generalized factor analysis; Integrative analysis; Network information; Bayesian factor analysis

Abstract

Rapid advances in technologies have led to the generation of massive amounts of -omics data. Integrative clustering of multiple omics data has received huge attention since it offers great promises in advancing biomedical research and clinical practice. The key idea of the integrative clustering is applying dimension reduction methods to multi-omics data for reducing the multi-high dimensionality so that clustering can be applied to the low-dimensional subspace holding most of the variations of the original data. This framework can identify disease subtypes or groups of related genes/cell lines from the same samples. In this work, we propose multi -omics integrative clustering analysis using a generalized Bayesian factor analysis (GFA) of multi-modal omics data incorporating biological information such as those from functional genomics or proteomics as a dimension reduction tool. We devise an EM algorithm for efficiently estimating the factor loadings in GFA. Our simulation studies demonstrate that the proposed method achieves better performance than existing methods. Real data application further demonstrates that the proposed method generates biologically meaningful results.