Genomic Ascertainment - Molecular and Genetic Aspects

National Human Genome Research Institute

Investigators

Linked publications & trials

Abstract

Infrastructure for the TGAC Project TGAC is The Genomic Ascertainment Cohort, a project to build capacity for genomic ascertainment research. A phase I cohort of 9,000 individuals has been entered into the study including ClinSeq participants, NIAID sequencing initiative participants, and Inova Hospital participants is now available to all NIH investigators at https://tgac.nhgri.nih.gov. Soon the nearly 5,000 NIEHS environmental polymorphism registry participants will be added. All of these individuals genomic data are available on the browser for viewing by intramural researchers. Then, they can apply to have us or their clinical team pursue phenotyping to test the hypothesis as to whether there is a phenotype associated with a given variant. We have already completed or have in process 20 such studies for intramural investigators. Two have been published or are in press. Based on its success, we have transferred this activity to become its own core facility within CPHR called the Reverse Phenotyping Core. Genomic Services Research Protocol (GSRP) A second activity of this project is the Genomic Services Research Protocol (GSRP). In this study, we are ascertaining individuals with secondary findings in one of the American College of Medical Genetics 59 gene secondary findings list. We have developed agreements or agreements in principle with 20 diagnostic laboratories that return secondary findings who refer patients to our study for evaluation. We intake individual and family histories and phenotypic evaluations. We supplement that by inviting selected participants to the NIHCC for deep phenotyping. We also perform molecular cascade testing to identify affected family members. Predictive genomic medicine or precision genomic medicine is of the highest medical importance and is an active area of research. We began piloting the return of genome-wide rare variants with the inception of the ClinSeq program in 2006, well before exome and genome sequencing became widespread (Biesecker et al., Genome Res 2009). We have continually expanded our work in this area by searching the genome for high penetrance variants that could be useful for clinical care, reproductive risks, pharmacogenetics, and other uses. My vision is that the genome is a health care resource, not a test, and that the cost of this health care resource can be amortized over the lifetime of an individual, long after its primary indicated usage is accomplished. Indeed, it was our work on ClinSeq that undergirded the recommendations we made for secondary findings (opportunistic screening) that were released in 2013 (Green et al., Genet Med 2013) These recommendations were initially highly controversial but with some amendments have now settled into accepted clinical practice an example of how we aim to change the practice of medicine. This paper has been cited more than 1,200 times and has also become the basis of a number of clinical testing laboratories health screening gene sets. ClinSeq has also been a leader in demonstrating to the field that genomes and exomes can be analyzed in healthy people and results returned, with minimal adverse consequences, dispelling the widespread fears about anxiety, depression, health care over-utilization, and other hypothetical risks. ClinSeq was also a charter member of the CSER (Clinical Sequencing Exploratory Research) consortium which allied six research centers to pilot genomics in health care settings. I joined CSER after participating in the study section that evaluated the CSER applications, reasoning that it would be beneficial for ClinSeq and for the other centers to pool efforts and experiences. Genomic medicine has at its foundation prediction predicting phenotype based on genotype. To address this foundational challenge, we need to develop research modes that can model and test the predictive power of genomic variants. Resources such as gnomAD provide critical data on population prevalence but are seriously limited with respect to phenotype data. The UK BioBank is supporting genotype ascertainment, but its data are limited to pre-hoc phenotypes and blood samples and participants cannot be re-contacted. There is thus a compelling need for the capacity to perform bespoke phenotyping based on genotypes. A spinoff opportunity from my CSER activities is the realm of variant interpretation, which is one of the main bottlenecks limiting the utility of exome and genome sequencing. We have developed a philosophy or approach that variant interpretation must be improved by orders of magnitude with respect to speed and cost, in order for the benefits of rare variant predictive medicine to be realized. To accomplish this, I have stepped up to co-chair the Sequence Variant Interpretation (SVI) working group of the ClinGen consortium. The primary function of the SVI working group was to clarify and refine the original ACMG/AMP pathogenicity criteria (Richards et al., Genet Med 2015). But we have, in addition, developed much larger goals which is to transition them from a subjective clinical practice into an objective, data-driven process that is amenable to at least partial automation. This ambitious goal is critical not only to the success of ClinGen, but to the success of predictive genomic medicine overall. In recognition of this leadership, I have been appointed to co-chair the ACMG/AMP/ClinGen working group to update the 2015 ACMG/AMP guidelines. In the realm of variant interpretation, our approach has been productive and fruitful with a number of advances that have supported the vision of objective, quantitative, data-driven variant interpretation. Some of the work included simple efforts, which was to jettison the reputable source criteria (PP5 & BP6) of the ACMG/AMP criteria (Richards et al., Genet Med 2015) in favor of using primary data (Biesecker et al., Genet Med 2018). We have also clarified and rationalized the PVS1 loss of function criterion (Abou Tayoun et al., Hum Mutat 2018) and the stand-alone criterion (BA1) for benign variants (Ghosh et al., Hum Mutat 2018). But the most significant advance I have made in leading SVI was to transition it from a categorical, combinatorial system to a quantitative, formally Bayesian foundation (Tavtigian et al., Genet Med 2018). In evaluating the Richards et al recommendations, I recognized that it was inherently structured as a nave Bayesian classifier, even though the original authors did not recognize it as such (and they explicitly set aside a quantitative or points-based framework). We developed a Bayesian framework and searched for actual values for prior and conditional probabilities that could be used to yield highly (but not completely) similar pathogenicity assertions as did Richards et al. This quantitative Bayesian framework now serves as the basis for many SVI discussions and activities such as reformulating the in silico criteria (PP3 & BP4) setting objective standards for functional data (PS3 & BS3), and to re-examine the linkage criteria (PP1 & BS4), which we have now recognized are logically entangled with the specific phenotype criterion (PP4). As well, we are collaborating with the Sanger institute to re-set the Bayesian prior probability to better reflect exome and genome testing, as the current prior is much too high for that testing modality. These scientific advances and these leadership roles (ClinGen SVI and ACMG/AMP/ClinGen) will position me to contribute significantly to the modernization of these recommendations toward the goal of automated variant interpretation, which addresses the critical throughput issue, as noted above. We have followed this up by publishing a points-based system for variant pathogenicity to simplify the process and have published Variant Curation Expert Panel Recommendations for RYR1.