An American-British team has developed a method for assessing how important mutations in specific noncoding regions of the genome are likely to be, and used it to look at nearly 100 cancer genomes to identify likely driver mutations in noncoding locations.
They published their results in the Oct. 4, 2013, issue of Science.
Interpreting the meaning of a mutation in a protein-coding sequence is hard enough. Even within the poster children of genetic risk, BRCA1 and BRCA2 genes, somewhere around 10 percent of mutations are “variants of unknown significance” that may or may not raise cancer risk.
Such challenges are magnified when the variants are in noncoding regions. While it has become clear over time that such regions are anything but the junk DNA they were once claimed to be, the function of noncoding DNA is still less well defined, especially since noncoding DNA describes what it does not do instead of what it does. Transcription factor binding sites and sites that are transcribed into RNA that is not translated, for example, have different roles despite the fact that they are both correctly considered noncoding regions of the genome.
The method first consisted of identifying noncoding regions of the genome that were under selective pressure by seeing how many rare variants it has. Variants in a genome region with a critical function are under strong selective pressure – if they improve that function they will rapidly spread through the population, and if they diminish it, they will just as rapidly be weeded out.
As a result, there are few common variants in important regions of the genome. Most of the variants there are rare ones, which have arisen fairly recently – and for that matter, are not that likely to stick around on an evolutionary timescale.
The authors first looked at data from both the 1 ,000 Genomes and the ENCODE projects to identify noncoding regions of the human genome that had mainly rare variants in them, much like protein-coding sequences do. (Despite the existence of mutations that have no functional consequences even within proteins, protein coding DNA sequences as a whole are under strong selective pressure.)
More detailed analysis of those regions showed that specific mutations within them were most likely to be disruptive: those that knocked out regions where proteins such as transcription factors were likely to bind, and those that sat in the center of a network, interacting with many other sites. To confirm the importance of sensitive and ultrasensitive regions to human health, the authors looked at another database, the human gene mutation database, and found that disease-causing mutations were more likely to be located in sensitive noncoding regions than insensitive ones.
Senior author Mark Gerstein, who is at Yale University, told BioWorld Today that the bigger goal of the work is to turn genomics data “into a practical tool, a workflow . . . that you can apply to a person who walks into the clinic.”
So in the second part of their study, Gerstein and his colleagues developed such a tool, and used it to apply their findings to almost 100 brain, breast and prostate cancer genomes.
Their methods identified nearly 100 mutations in noncoding regions that were nevertheless likely to be so-called driver mutations – mutations that are a cause and not just a consequence of cancer.
Gerstein is optimistic that such mutations in noncoding regions can be therapeutically targeted. While doing so is overall less straightforward than developing a kinase inhibitor for a mutated protein, “I don’t think [noncoding regions] are not druggable,” he said.
And given the statistics, focusing on the 1 percent of mutations that are in coding regions is like looking under the streetlight for a lost object. “Most cancer mutations – about 99 percent of them – are in noncoding regions,” he said. “It’s foolish not to look at them.”