Evaluation of integrative clustering methods for the analysis of multi-omics data

Abstract

Recent advances in sequencing, mass spectrometry and cytometry technologies have enabled researchers to collect large-scale omics data from the same set of biological samples. The joint analysis of multiple omics offers the opportunity to uncover coordinated cellular processes acting across different omic layers. In this work, we present a thorough comparison of a selection of recent integrative clustering approaches, including Bayesian (BCC and MDI) and matrix factorization approaches (iCluster, moCluster, JIVE and iNMF). Based on simulations, the methods were evaluated on their sensitivity and their ability to recover both the correct number of clusters and the simulated clustering at the common and data-specific levels. Standard non-integrative approaches were also included to quantify the added value of integrative methods. For most matrix factorization methods and one Bayesian approach (BCC), the shared and specific structures were successfully recovered with high and moderate accuracy, respectively. An opposite behavior was observed on non-integrative approaches, i.e. high performances on specific structures only. Finally, we applied the methods on the Cancer Genome Atlas breast cancer data set to check whether results based on experimental data were consistent with those obtained in the simulations.

Benchmarking joint multi-omics dimensionality reduction approaches for cancer study

“…We benchmarked in-depth nine jDR algorithms, representative of multi-omics integration approaches, in the context of cancer data analysis. In contrast to existing comparisons [10][11][12][13] , our benchmark not only focuses on the evaluation of the clustering outputs, but also evaluates the biological, clinical, and survival annotations of the factors and metagenes. Existing comparisons also mainly use simulated data; we here consider large datasets of bulk cancer multi-omics as well as single-cell data.…”

“…Dimensionality Reduction (DR) approaches decompose the omics into a shared low-dimensional latent space 8,9 . Four recent reviews tested and discussed some of these existing methods from the clustering performance perspective [10][11][12][13] . Pierre-Jean et al 12 , Rappoport et al 10 and Tini et al 13 selected one method from each of the aforementioned three categories, while Chauvel et al 11 focused on Bayesian and DR approaches.…”

“…Thus, DR simultaneously provides information on all the key objectives mentioned above, namely the classification of samples into subtypes, their association with outcome/survival as well as the reconstruction of their underlying molecular mechanisms. As a consequence, the design of DR approaches for the joint analysis and integration of multiple omics (jDR) is currently a highly active area of research 5,8,10,11,14 .…”

Progress and applications of mass cytometry in sketching immune landscapes

“…49,50 CyTOF can also work in conjunction with other techniques including single-cell genome and transcriptome profiling 169,170 and bioinformatic pipelines. 171,172 For instance, Zheng et al conducted CyTOF profiling of immune microenvironment in hepatocellular carcinoma and revealed that leading-edge regions exhibited an increase of tumor-associated CD4/CD8 double-positive T (DPT) cells, which synergistically expressed PD-1/HLA-DR/ICOS/CD45RO. 173 Single-cell RNA-seq was employed to characterize DPT cells and specifically identified PD-1 high DPT cluster derived from intratumoral CD8 + T cells.…”

Heterogeneous Multi-Layered Network Model for Omics Data Integration and Analysis

“…In this review, we focus on the cross-domain relation inference (or link prediction) problem for the HMLN. Readers can refer other excellent reviews of the multiplex networks (Chauvel et al, 2019).…”