Integrative analysis of a cancer somatic mutome

Abstract

Background: The consecutive acquisition of genetic alterations characterizes neoplastic processes. As a consequence of these alterations, molecular interactions are reprogrammed in the context of highly connected and regulated cellular networks. The recent identification of the collection of somatically mutated genes in breast tumors (breast cancer somatic "mutome") allows the comprehensive study of its function and organization in complex networks.

Results: We analyzed functional genomic data (loss of heterozygosity, copy number variation and gene expression in breast tumors) and protein binary interactions from public repositories to identify potential novel components of neoplastic processes, the functional relationships between them, and to examine their coordinated function in breast cancer pathogenesis. This analysis identified candidate tumor suppressors and oncogenes, and new genes whose expression level predicts survival rate in breast cancer patients. Mutome network modeling using different types of pathological and healthy functional relationships unveils functional modules significantly enriched in genes or proteins (genes/proteins) with related biological process Gene Ontology terms and containing known breast cancer-related genes/proteins.

Conclusion: This study presents a comprehensive analysis of the breast somatic mutome, highlighting those genes with a higher probability of playing a determinant role in tumorigenesis and better defining molecular interactions related to the neoplastic process.

Figure 1

Integration of LOH, CN and expression data to better define candidate tumor suppressors and oncogenes for the breast cancer neoplastic process. Examples of LOH and CN analyses: (A) LOH analysis for HSA1 shows three critical regions (defined by close boundaries of LOH) indicated by pink lines across tumor samples; (B) CN analyses indicate GAB1 locus genomic amplification in HSA4, and SORL1 and TECTA loci genomic loss in HSA11; and (C) Integration of LOH and CN, and differential expression in tumors relative to healthy tissues indicate candidate tumor suppressors (down-regulated in tumors, green) and oncogenes (up-regulated in tumors, red) in four different types of breast tumors as indicated by numbers in brackets.

Figure 2

Gene co-expression analysis in breast tumors. Clustering of microarray probes (297 × 297) representing mutome (validated and non-validated) [1] and benchmark (literature) [9] genes according to absolute PCC values. Clusters are named according to the benchmark(s) gene(s) present in each of them (i.e. RB1, ETV6-NTKR3 or TP53-related). Boxes contain validated mutome genes present in each cluster. Non-validated gene names are not shown.

Figure 3

Gene expression analysis and breast cancer survival. Kaplan-Meier survival curves based upon categorized expression in tertiles are shown for three validated genes in the Hu et al. [13] data set.

Figure 4

Human interactome network analysis, functional prediction and breast and colorectal cancer mutome association. (A) Predicted interactions for uncharacterized validated mutome gene products. Functional assignment is based on GO term annotations. Protein interactions and node types are indicated as shown in the insets. (B) Breast and colorectal cancer mutome association through extracellular matrix and cytoskeleton constituents.

Figure 5

Human interactome network analysis, direct interactions between mutome gene products. (A) Left panel, direct interactions between validated mutome and/or benchmark gene products. Right panel, interactions centered on SPTAN1, whose expression level predicts survival (Fig. 3). Grey nodes represent non-mutome/benchmark proteins. (B) Network generated by direct protein interactions between validated and non-validated mutome and/or benchmark gene products (top left inset). An image of the largest component of this network is shown, with critical nodes that connect benchmark or mutome proteins indicated by arrows. (C) Clusters or densely connected regions in the interactome network that contain more mutome gene products than expected by chance: cluster A shows enrichment in annotations of the TGF-beta and insulin signaling pathways and of DNA transcriptional activity; cluster B shows enrichment for centrosome-related tasks and DNA transcriptional activity.

Figure 6

Breast cancer mutome network modeling. Left panel, five functional genomic or proteomic, pathological or healthy-related associations; each one indicated by one of the colored lines shown in the inset was included to generate a mutome network model. Right panel, clusters or densely connected regions in the network that show enrichment in GO terms (functional modules). Benchmark nodes present in these functional modules are marked by arrows.