Characterizing genomic alterations in cancer by complementary functional associations

Abstract

Systematic efforts to sequence the cancer genome have identified large numbers of relevant mutations and copy number alterations in human cancers; however, elucidating their functional consequences, and their interactions to drive or maintain oncogenic states, is still a significant challenge. Here we introduce REVEALER, a computational method that identifies combinations of mutually exclusive genomic alterations correlated with functional phenotypes, such as the activation or gene-dependency of oncogenic pathways or the sensitivity to a drug treatment. We use REVEALER to uncover complementary genomic alterations associated with the transcriptional activation of β-catenin and NRF2, MEK-inhibitor sensitivity, and KRAS dependency. REVEALER successfully identified both known and new associations demonstrating the power of combining functional profiles with extensive characterization of genomic alterations in cancer genomes.

Kataegis Expression Signature in Breast Cancer Is Associated with Late Onset, Better Prognosis, and Higher HER2 Levels

“…To investigate the molecular underpinnings of the clinical differences between tumors with and those without kataegis loci, we used the information coefficient (IC) (Kim et al, 2016) to examine the 97 TCGA tumors with known kataegis status. IC is an information-theoretic association metric that allows the discovery of linear or non-linear correlations between genomic alterations (here the presence or absence of kataegis loci) and functional phenotypes, such as gene expression levels, protein expression data, and pathway enrichment (Table S5).…”

“…The IC analysis (Kim et al, 2016) was performed using 11 distinct datasets as input: (1) RNA-seq gene expression levels derived from TCGA (19,921 genes), (2) reverse-phase protein array (RPPA) levels for 142 proteins derived from TCGA, (3) ssGSEA pathway enrichment levels for 4,496 chemical and genetic perturbation gene sets derived from MSigDB (Liberzon et al, 2011, 2015), (4) ssGSEA pathway enrichment levels for 217 Biocarta pathways, (5) ssGSEA pathway enrichment levels for 674 Reactome pathways, (6) ssGSEA enrichment levels for the targets of 221 microRNA, (7) ssGSEA enrichment levels for the targets of 608 transcription factors, (8) ssGSEA enrichment levels for 426 cancer gene neighborhoods, (9) ssGSEA enrichment levels for 431 cancer modules, (10) ssGSEA enrichment levels for 279 oncogenic signatures, and (11) ssGSEA enrichment levels for 52 hallmark gene sets derived from MSigDB (Liberzon et al, 2011, 2015). The IC analysis was run separately for (1) all tumors with kataegis, (2) tumors with kataegis loci on chromosome 8, (3) tumors with kataegis loci on chromosome 17, and (4) tumors with kataegis loci on chromosome 22.…”

“…The IC analysis was run separately for (1) all tumors with kataegis, (2) tumors with kataegis loci on chromosome 8, (3) tumors with kataegis loci on chromosome 17, and (4) tumors with kataegis loci on chromosome 22. The IC analysis is based on the use of a normalized differential mutual information function, which is a non-linear correlation coefficient (Kim et al, 2016) with values in the range [−1,1], with 1 representing a perfect match with kataegis and −1 representing a perfect anti-match. For the cases of binary phenotypes (e.g., the kataegis status) and continuous genomic features (e.g., gene expression), the IC becomes the generalized Jensen-Shannon divergence (Kim et al, 2016).…”

Precision Oncology: The Road Ahead

“…One group of methods searches for significantly mutated groups of genes in known pathway databases [97] and protein interaction networks [35, 93, 98–100]. Other methods search for functional mutations that co-occur with sample-level events [101, 102], which can be viewed as a supervised learning task. These approaches have been used on cancer cell line drug sensitivity and gene dependency/addiction (i.e.…”

“…These include applying various machine learning methodologies to predict genetic interactions in different species [128–131], and in cancer (employing yeast SLi) [119, 132], utilizing metabolic modeling [133, 134], evolutionary characteristics [119, 129], transcriptomic profiles [101, 135], and more recently, by mining cancer patient data [136–138] (Table S2D). One recent study evaluated the TCGA copy number and transcriptomics data to identify, as candidate SLis, gene pairs that are almost never found inactivated in the same tumors [136].…”