Integrative clustering of multiple genomic data types using a joint latent variable model with application to breast and lung cancer subtype analysis

TLDR

Tumor molecular complexity spans genomic, epigenomic, transcriptomic and proteomic levels, yet effective integrative clustering tools are scarce and the prevailing method relies on separate clustering followed by manual integration. The study aims to develop a joint latent variable model that simultaneously integrates multiple genomic data types to enable a unified clustering of tumor subtypes. iCluster models the joint distribution of multiple data types with a latent variable framework, flexibly captures inter‑type associations and intra‑type covariance, reduces dimensionality, and estimates parameters via an EM algorithm, as illustrated on breast and lung cancer copy‑number and expression data. The algorithm automatically identified concordant DNA copy‑number and expression subtypes, as well as unique profiles specific to each data type, and revealed potentially novel subtypes by integrating weak yet consistent alteration patterns. R code for iCluster is available at http://www.mskcc.org/mskcc/html/85130.cfm.

Abstract

The molecular complexity of a tumor manifests itself at the genomic, epigenomic, transcriptomic and proteomic levels. Genomic profiling at these multiple levels should allow an integrated characterization of tumor etiology. However, there is a shortage of effective statistical and bioinformatic tools for truly integrative data analysis. The standard approach to integrative clustering is separate clustering followed by manual integration. A more statistically powerful approach would incorporate all data types simultaneously and generate a single integrated cluster assignment.We developed a joint latent variable model for integrative clustering. We call the resulting methodology iCluster. iCluster incorporates flexible modeling of the associations between different data types and the variance-covariance structure within data types in a single framework, while simultaneously reducing the dimensionality of the datasets. Likelihood-based inference is obtained through the Expectation-Maximization algorithm.We demonstrate the iCluster algorithm using two examples of joint analysis of copy number and gene expression data, one from breast cancer and one from lung cancer. In both cases, we identified subtypes characterized by concordant DNA copy number changes and gene expression as well as unique profiles specific to one or the other in a completely automated fashion. In addition, the algorithm discovers potentially novel subtypes by combining weak yet consistent alteration patterns across data types.R code to implement iCluster can be downloaded at http://www.mskcc.org/mskcc/html/85130.cfm