In the world of cancer genomics, mutational hotspots and cancer drivers are used somewhat interchangeably.
"It was common knowledge... that if you see a site [mutated] in many cancers, that site is a driver," Gad Getz told BioWorld.
But Getz, who is director of the Cancer Genome Computational Analysis Group at the Broad Institute of MIT and Harvard, and his team argue that the conventional wisdom is incorrect.
In the Sept. 16, 2019, issue of Cancer Cell, they described developing a mathematical model that takes into account the fact that not all regions of the genome have equal mutation rates.
Taking into account the different rates showed that "many genes are totally neutral, but still have these hotspots," Getz said.
First author Julian Hess, computational scientist at the Broad Institute, told BioWorld that the model is meant to help eliminate false positives in the hunt for driver mutations.
"The whole initial impetus for this project was that as these [sequenced] patient cohorts were getting larger and larger, the conventional methods were nominating far too many drivers to be plausible," Hess said.
In some highly mutated tumor types, using hotspot mutations as a proxy for drivers pointed to the existence of 100 or more drivers.
Getz said that eliminating false positives is an important way to streamline the hunt for actionable mutations.
"People study cancer genomes with the goal of finding driver genes, and driver alleles within those genes," he said. But to get from a potential driver to a validated therapeutic target takes "a lot of activities that ... are time-consuming and expensive and tedious."
In other words, the sooner false positives are eliminated, the better – another instance of the virtues of failing early.
How genomes act locally
On the scientific side, the work shows the importance of genomic features that affect the likelihood of mutations in very specific genomic locations.
"We've known for the better part of a decade that some features [affect] mutability on the scale of the gene," Hess said.
Later replicating parts of the genome are more mutable than average, for example, while highly expressed genes are less mutable because of the existence of transcription coupled-repaired mechanisms.
More recently, it has become clear that there are also genomic features that affect mutability in a much more localized fashion – "on the scale of a few nucleotides," Hess said.
Earlier this year, co-corresponding author Michael Lawrence and his colleagues at Massachusetts General Hospital published a paper describing one such feature. In the APOBEC family of enzymes, they showed that parts of the APOBEC genes are prone to folding into hairpins, and that such hairpins have an increased likelihood of accumulating mutations.
APOBEC enzymes are up-regulated in many cancer cells, and so the work is an example of understanding the a priori likelihood of a mutation occurring at the "mesoscale" of about 30 nucleotides. Driver mutations in APOBEC enzymes do occur, but they are outside of the regions that form hairpin loops.
"We suspect there will be many more features like this Apobec hairpin," Hess said.
Currently, the model "can quantify the variability that is explained by known features, and the amount that has yet to be explained," he added.
Doing so, Getz added, has shown that there remains "a lot of unexplained variability" in mutation rates.
He predicted that "additional genomic features will be found in the future" to explain that variability.