Biomolecular events in cancer revealed by attractor metagenes

TL;DR

This paper introduces an iterative, seed-based method to identify core biomolecular events in cancer through attractor metagenes, revealing common and specific gene expression patterns across multiple datasets.

Contribution

The authors present a novel unconstrained approach for generating attractor metagenes that highlight underlying biological mechanisms in cancer.

Findings

Identified multiple biomolecular events across cancer types

Discovered common and unique gene expression patterns

Revealed biological processes like mesenchymal transition and chromosomal instability

Abstract

Mining gene expression profiles has proven valuable for identifying metagenes, defined as linear combinations of individual genes, serving as surrogates of biological phenotypes. Typically, such metagenes are jointly generated as the result of an optimization process for dimensionality reduction. Here we present an unconstrained method for individually generating metagenes that can point to the core of the underlying biological mechanisms. We use an iterative process that starts from any seed gene and converges to one of several precise attractor metagenes representing biomolecular events, such as cell transdifferentiation or the presence of an amplicon. By analyzing six rich gene expression datasets from three different cancer types, we identified many such biomolecular events, some of which are present in all tested cancer types. We focus on several such events including a stage-associated mesenchymal transition and a grade-associated mitotic chromosomal instability.