Are sites with multiple single nucleotide variants in cancer genomes a consequence of drivers, hypermutable sites or sequencing errors?

Hotspot propensity across mutational processes

The sparsity of mutations observed across tumours hinders our ability to study mutation rate variability at nucleotide resolution. To circumvent this, here we investigated the propensity of mutational processes to form mutational hotspots as a readout of their mutation rate variability at single base resolution. Mutational signatures 1 and 17 have the highest hotspot propensity (5-78 times higher than other processes). After accounting for trinucleotide mutational probabilities, sequence composition and mutational heterogeneity at 10 Kbp, most (94-95%) signature 17 hotspots remain unexplained, suggesting a significant role of local genomic features. For signature 1, the inclusion of genome-wide distribution of methylated CpG sites into models can explain most (80-100%) of the hotspot propensity. There is an increased hotspot propensity of signature 1 in normal tissues and de novo germline mutations. We demonstrate that hotspot propensity is a useful readout to assess the accuracy of mutation rate models at nucleotide resolution. This new approach and the findings derived from it open up new avenues for a range of somatic and germline studies investigating and modelling mutagenesis.

The landscape and driver potential of site-specific hotspots across cancer genomes

Large sets of whole cancer genomes make it possible to study mutation hotspots genome-wide. Here we detect, categorize, and characterize site-specific hotspots using 2279 whole cancer genomes from the Pan-Cancer Analysis of Whole Genomes project and provide a resource of annotated hotspots genome-wide. We investigate the excess of hotspots in both protein-coding and gene regulatory regions and develop measures of positive selection and functional impact for individual hotspots. Using cancer allele fractions, expression aberrations, mutational signatures, and a variety of genomic features, such as potential gain or loss of transcription factor binding sites, we annotate and prioritize all highly mutated hotspots. Genome-wide we find more high-frequency SNV and indel hotspots than expected given mutational background models. Protein-coding regions are generally enriched for SNV hotspots compared to other regions. Gene regulatory hotspots show enrichment of potential same-patient second-hit missense mutations, consistent with enrichment of hotspot driver mutations compared to singletons. For protein-coding regions, splice-sites, promoters, and enhancers, we see an excess of hotspots associated with cancer genes. Interestingly, missense hotspot mutations in tumor suppressors are associated with elevated expression, suggesting localized amino-acid changes with functional impact. For individual non-coding hotspots, only a small number show clear signs of positive selection, including known sites in the TERT promoter and the 5' UTR of TP53. Most of the new candidates have few mutations and limited driver evidence. However, a hotspot in an enhancer of the oncogene POU2AF1, which may create a transcription factor binding site, presents multiple lines of driver-consistent evidence.

Computational methods for detecting cancer hotspots

Cancer mutations that are recurrently observed among patients are known as hotspots. Hotspots are highly relevant because they are, presumably, likely functional. Known hotspots in BRAF, PIK3CA, TP53, KRAS, IDH1 support this idea. However, hundreds of hotspots have never been validated experimentally. The detection of hotspots nevertheless is challenging because background mutations obscure their statistical and computational identification. Although several algorithms have been applied to identify hotspots, they have not been reviewed before. Thus, in this mini-review, we summarize more than 40 computational methods applied to detect cancer hotspots in coding and non-coding DNA. We first organize the methods in cluster-based, 3D, position-specific, and miscellaneous to provide a general overview. Then, we describe their embed procedures, implementations, variations, and differences. Finally, we discuss some advantages, provide some ideas for future developments, and mention opportunities such as application to viral integrations, translocations, and epigenetics.

Passenger Hotspot Mutations in Cancer

Current statistical models for assessing hotspot significance do not properly account for variation in site-specific mutability, thereby yielding many false-positives. We thus (i) detail a Log-normal-Poisson (LNP) background model that accounts for this variability in a manner consistent with models of mutagenesis; (ii) use it to show that passenger hotspots arise from all common mutational processes; and (iii) apply it to a ∼10,000-patient cohort to nominate driver hotspots with far fewer false-positives compared with conventional methods. Overall, we show that many cancer hotspot mutations recurring at the same genomic site across multiple tumors are actually passenger events, recurring at inherently mutable genomic sites under no positive selection.