NEW YORK – An international team led by investigators at Harvard University has found that inherited common genetic variants, including those peppered across non-coding parts of the genome, come together to influence cancer risk by impacting the expression of genes involved in a handful of key processes.
"Genome-wide association studies have analyzed hundreds of thousands of individuals to find genetic variants that are associated with increased risk of developing cancer, but many of these fall in intergenic regions and have no clear functional association with cancer drivers," senior author John Quackenbush, a biostatistics, computational biology, and cancer biology researcher affiliated with Brigham and Women's Hospital, the Dana-Farber Cancer Institute, and Harvard TH Chan School of Public Health, and his co-authors wrote in a paper published online yesterday in the British Journal of Cancer.
In an effort to understand potential ties to cancer risk for 872 SNPs implicated in prior genome-wide association studies of susceptibility for traits and conditions related to 41 kinds of cancer, Quackenbush and his colleagues used a systems biology-based strategy to try to understand the roles for these variants, which appeared far more likely to have potential regulatory roles than to affect protein-coding or splicing sites in the genome.
With that in mind, the team began by putting together so-called "bipartite" networks of expression quantitative trait loci (eQTLs) that are active within or across more than a dozen tissue types using matched RNA sequence and genotyping data from the Genotype-Tissue Expression (GTEx) effort.
After identifying the types of genes and SNPs that made up shared and distinct eQTL networks in the 12 primary human tissues and one cell line considered, the investigators overlaid SNP clues from cancer susceptibility GWAS, uncovering apparent connections between cancer-related SNPs and networks containing immune genes, including major histocompatibility complex class I and class II genes, and genes with tissue-specific roles.
Along with an over-representation in regulatory sites for oncogenes or tumor suppressor genes, they noted that non-coding SNPs associated with cancer risk were also more prone to fall at central sites in these networks, suggesting they may influence the expression and regulation of many genes.
"By mapping cancer risk SNPs to bipartite networks built from both cis- and trans-eQTLs in thirteen tissues, we show that cancer risk SNPs play a distinctive role in defining the structure of such networks," the authors reported. "Cancer risk SNPs are associated not only with cancer genes, but with many other genes associated with biological functions that can be linked to cancer development and progression."
The researchers speculated that such findings may ultimately inform efforts to predict individuals' cancer risk using artificial intelligence models that take into account information at many, many variants per person.
"[O]ur results show that small genetic variations work collectively to subtly shift the activity of genes that drive cancer," Quackenbush said in a statement. "We hope that this approach could one day save lives by helping to identify people at risk of cancer, as well as other complex diseases."