Skip to main content
Premium Trial:

Request an Annual Quote

New Statistical Model Identifies Rare Tumor Suppressing Genes

NEW YORK (GenomeWeb) – Using a pan-cancer analysis called allele-specific copy number analysis of tumors (ASCAT), researchers at the Francis Crick Institute, the University of Leuven, and their colleagues developed a new type statistical model, which they were able to use to identify 27 new tumor suppressing genes.    

For their study, which was published today in Nature Communications, the researchers screened for homozygous deletions in 2,218 tumors from 12 cancer types including breast, lung, and bowel cancers, in an attempt to identify rare tumor suppressors. They identified 96 genomic regions that were recurrently targeted by homozygous deletions, either over tumor suppressors or over fragile sites, and constructed a statistical model that separated fragile sites from regions showing signatures of positive selection for homozygous deletions in order to identify candidate tumour suppressors within those regions.

"We find 16 established tumor suppressors and propose 27 candidate tumor suppressors," the authors wrote. "Several of these genes (including MGMT, RAD17, and USP44) show prior evidence of a tumor suppressive function. Other candidate tumor suppressors, such as MAFTRR, KIAA1551, and IGF2BP2, are novel. Our study demonstrates how rare tumor suppressors can be identified through copy number meta-analysis."

The researchers began by building a compendium of 2,218 publically available primary tumor samples hybridized to Affymetrix 250K StyI SNP arrays, and used ASCAT to infer tumor purity and ploidy and derive copy number profiles. They then performed a systematic screen for homozygous deletions across the 2,137 primary tumors that passed ASCAT analysis, and identified 1,865 homozygous deletions in total.

They found that the genomic distribution of homozygous deletions was skewed towards specific regions, and devised two distinct permutation tests in order to investigate this further. In the first experiment, they modeled homozygous deletions as a combination of two independent events. "Based on this test, homozygous deletions, and particularly large homozygous deletions, are strongly depleted across the genome," the team wrote. "This paucity of homozygous deletions is not unexpected, and may be ascribed to negative selection: removal of both copies of any functional gene or other element in the genome likely results in a selective disadvantage for the cell."

In the second experiment, they permuted homozygous deletions as singular events, keeping the total homozygously deleted sequence constant for each sample across permutations. This caused a total of 42.6 megabases of the genome, distributed across 96 distinct regions, to be targeted more frequently by homozygous deletions than expected.

"We observed 15 peaks of homozygous deletions over 16 established tumor suppressors," the authors added. They found that CDKN2A was the dominant homozygously deleted tumor suppressor, with 108 homozygous deletions across nine cancer types.

The researchers also found that homozygous deletions over tumor suppressors are enriched due to positive selection, but that homozygous deletions over fragile sites are enriched due to a local increase in genomic instability. Therefore, the structural signature of deletions is distinct in fragile sites, compared to regions harbouring tumor suppressors.

"Small hemizygous deletions reflect local fragility, as they require two DNA breakage events in close proximity. Large hemizygous deletions, in contrast, are due to several other mechanisms, such as whole-chromosome or whole-arm loss," they wrote. "As a result, fragile sites are characterised by frequent small deletion events, while regions containing tumor suppressors more frequently show large deletion events. This difference was exploited to construct three metrics that discriminate fragile sites from regions containing tumor suppressors."

By looking for these differences, the researchers were able to construct a statistical model that is able to reject local fragility for 13 of the 15 peaks containing known tumor suppressors, with the exception of VHL/FANCD2 and CDKN2C. In addition, the model identified 32 regions that are unlikely to be explained by local fragility, showing a signature of positive selection.

"We overlaid the patterns of homozygous deletions in these regions with mutation data from COSMIC and with scientific literature, aiming to identify candidate tumor suppressors," the authors wrote. "Two regions were intergenic, and for four additional regions the targets remain unknown. In each of the remaining 26 regions, we were able to pinpoint at least one candidate tumor suppressor. Our proposed tumor suppressors include genes previously suggested to play a role in oncogenesis (e.g., MGMT and USP44), as well as novel candidates (e.g., KIAA1551, CASP3, and MAFTRR).

Further, the researchers noted that their results provided a complementary view of tumor suppressor genes to sequencing screens for recurrent single-nucleotide substitutions and small insertions and deletions. "We conjecture that this study identifies a class of predominantly rare tumor suppressors, such as CPEB3 and MGMT, that are more prone to be inactivated by homozygous deletions than point mutations, a proportion of which therefore may not be readily identifiable through mutation analysis given current sample sizes," they added.