Skip to main content
Premium Trial:

Request an Annual Quote

Pan-Cancer Study Points to Gene Networks Prone to Rare Mutations in TCGA Tumors

NEW YORK (GenomeWeb) – A cross-cancer analysis appearing online today in Nature Genetics used data from thousands of tumors assessed through the Cancer Genome Atlas program to identify sub-networks of genes that tend to be altered in one or more cancer types.

Researchers from Brown University and other centers in the US, Spain, and Denmark used a newly developed computation method to look for sub-networks of genes that tended to carry mutations and/or copy number changes in a dozen cancer types considered by TCGA. Their search led to 16 sub-networks that spanned pathways with new and known roles in cancer.

Along with components of pathways involved in signaling pathways, chromatin remodeling, and cell cycle regulation, for example, the team narrowed in on gene sub-networks involved in chromosome organization and cohesion. While some sub-networks tended to be mutated in particular cancers, others were identified owing to rare mutations that turned up across multiple types of cancer.

"The next step is translating all of this information from cancer sequencing into clinically actionable decisions," corresponding author Benjamin Raphael, a computer science researcher at Brown University, said in a statement. 

"For example, there are now drugs that are used to treat patients who have mutations in particular genes. However, perhaps patients who don't have a mutation in the targeted gene, but have a mutation in the same pathway, might respond to the same drug," Raphael explained, noting that more research is needed to explore such possibilities.

Sequencing studies done in recent years have unearthed a vast array of genes that can be mutated in cancer, Raphael and colleagues noted. But only a relatively small subset contains mutations in a statistically significant number of tumors from one or more cancer types. That has left scientists puzzling over the best way to parse out meaningful low-frequency mutations from mutations that merely occur as passenger events.

Studies that consider mutation frequency for genes with recognized roles in shared molecular pathways — including past work by the pan cancer arm of TCGA — have started to define oft-mutated pathways within the mishmash of passenger mutations. So far, though, such analyses have been limited to picking frequent mutations within genes from relatively well-defined pathways.

"Although statistical tests of enrichment in known pathways or gene sets exist, such tests do not identify new pathways, have limited power to evaluate cross-talk between pathways, and generally ignore the topology of interactions between genes," authors of the new study wrote.

In an effort to get a broader look at gene networks that may contribute to various cancer types, the team used an algorithm known as HotNet2 to scrutinize exome sequences, RNA sequences, array-based copy number profiles, and other types of data available for 3,281 tumors from 12 cancer types tested by TCGA.

The HotNet2 algorithm is intended to consider interactions amongst protein encoded by given genes, as well as the "heat" emitted by proteins at highly connected hub sites within networks and the direction that it moves in relation to other spokes of the network.

As such, the study's authors argued that the analytical approach "delves deeper into the long tail of rarely mutated genes and also assembles combinations of individual genes into a relatively small number of interacting networks."

Once they tossed out data for samples with rampant mutations and genes with muted expression across all of the samples, the researchers were left with data for 3,110 tumors and more than 11,500 genes.

After estimating gene "heat" with the help of existing network interaction data and mutation frequencies found in tumors, the researchers used HotNet2 to identify 14 mutated gene sub-networks. They added in two additional sub-networks to that, including a sub-network containing the condensin complex pathway, based on individual interaction networks.

Well-known cancer culprits such as the TP53 gene were among the sub-networks detected, as were NOTCH and PI3 kinase signaling pathways.

These three sub-networks were not only frequently mutated (almost 82 percent of tumors carried mutations in at least one of them), but also highly interconnected by linker genes, researchers reported. And their analytical algorithm highlighted interactions between these genes and players not previously implicated in cancer.

Other sub-networks contained genes from chromatin-remodeling complexes such as SWI/SNF or BAP1, as well as genes contributing to cohesion between sister chromatids during cell division. 

Some sub-networks seemed susceptible to being mutated together in a given tumor, while other sub-network mutations were mutually exclusive. Similarly, the team saw examples of sub-networks with an over-representation of mutations in specific cancer types and others that were spread out across cancer types.

Those involved in the study noted that such mutation networks will likely become larger, but more finely curated, as researchers develop a clearer understanding of how proteins interact in various tissues in general and as tumor datasets expand to include additional information on epigenetic and structural variations detected in cancer.

The team developed an interactive, online viewer for those interested in exploring the pan-cancers sub-networks in more detail.