Multi-omic and multi-view clustering algorithms: review and cancer benchmark

By Nimrod Rappoport, Ron Shamir. bioRxiv, 2018

440 citations recorded by the source corpus.

High throughput experimental methods developed in recent years have been used to collect large biomedical omics datasets. Clustering of such datasets has proven invaluable for biological and medical research, and helped reveal structure in data from several domains. Such analysis is often based on investigation of a single omic. The decreasing cost and development of additional high throughput methods now enable measurement of multi-omic data. Clustering multi-omic data has the potential to reveal further systems-level insights, but raises computational and biological challenges. Here we review algorithms for multi-omics clustering, and discuss key issues in applying these algorithms. Our review covers methods developed specifically for multi-omic data as well as generic multi-view methods developed in the machine learning community for joint clustering of multiple data types. In addition, using cancer data from TCGA, we perform an extensive benchmark spanning ten different cancer types, providing the first systematic benchmark comparison of leading multi-omics and multiview clustering algorithms. The results highlight several key questions regarding the use of single-vs. multi-omics, the choice of clustering strategy, the power of generic multi-view methods and the use of approximated p-values for gauging solution quality. Due to the rapidly increasing use of multi-omics data, these issues may be important for future progress in the field.

Subfields: Information Systems, Molecular Biology

View Paper: https://www.semanticscholar.org/paper/7953e9ad29893cf8320134a4567e5382f1545a3d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Information Systems

In Molecular Biology

Citations in the corpus

Integration strategies of multi-omics data for machine learning analysis (2021, background): “With the notable exception of Herrmann et al. (2020) [168] which focused on survival prediction methods for multi-omics data, most benchmarks focus on clustering and dimensionality reduction methods [14,26,27,135,169–171].”
Uncovering Prognosis-Related Genes and Pathways by Multi-Omics Analysis in Lung Cancer (2020, result): “Clustering analyses applied with clinic lly associat d f atures demonstrated that K = 2 was the optimal number, and this is concordant with the previous report performing with t 10 TCGA cancer dataset [49].”
A guide to machine learning for biologists (2021, background): “3e) is pervasive within biology 42,43 . k - means is a strong general purpose approach to clustering that, like many other clustering algorithms, requires the number of clusters to be set as a hyperparameter 44 .”
The Need for Multi-Omics Biomarker Signatures in Precision Medicine (2019, background): “Analytical tools have been suggested for these types of complex datasets, such as clustering approaches [75], or weighted gene co-expression network analysis (WGCNA, [76]).”