The commoditization of exome and genome sequencing is reshaping basic research and clinical practice in genetics. Despite successes in implicating casual disease loci through large-scale sequencing, there are no similarly straightforward means to distinguish between pathogenic and neutral alleles for a given gene. Even for genes with long-standing disease associations, a substantial fraction of clinically observed alleles are classified as ?variants of unknown significance?, or VUS. For instance, mutations in DNA mismatch repair (MMR) genes strongly predispose to colorectal and other cancers and are frequently screened for in the clinic, yet >30% of all known MMR gene mutations are classified as VUS (Thompson et al., Nature Genetics, 2013). This uncertainty poses a considerable obstacle to the goal of genotype-driven treatment. We propose to address this challenge by developing scalable technologies to synthesize and functionally screen mutations in a massively parallel fashion. We anticipate that these tools will be broadly useful, and we will apply them first to human MMR gene mutations, with the following specific aims: (1) to profile the functional activity of every possible missense variant for three human MMR genes (MSH2, MLH1, and MSH6), and to predict the resulting likelihood of pathogenicity across alleles. We will construct comprehensive allelic series for these genes, introduce them into relevant MMR-null mammalian tissue culture models, select en masse for restoration of MMR activity, and count the resulting allelic depletion by deep sequencing; (2) to survey the prevalence and degree of epistatic effects between and within MMR genes by introducing libraries of double mutants and screening as in Aim 1; and, (3) to directly characterize the types and frequencies of mutations caused by inactivation of MMR using whole-genome sequencing in isogenic cell lines following repeated passaging. Completion of these aims will clarify sequence-structure-function relationships for human MMR genes, and will allow derivation of a risk score for clinically observed mutations. The methodological advances proposed here will provide a foundation for generalized approaches to dissect the allelic heterogeneity of human oligogeneic disorders, and a path toward functional annotation of the rapidly growing VUS catalogs.
Increasingly many genes have been identified as risk factors for cancer or other disorders, yet diagnostic sequencing of these genes often produces uncertain results. This poses a growing problem as the use of clinical targeted and whole-genome sequencing expands. Improved technologies for creating and testing very large sets of mutations will help to establish which specific mutations cause disease, and which are benign.
