Paul A. Scheet - US grants

DESCRIPTION (provided by applicant): Massively-parallel ("next-generation") shotgun DNA sequencing projects will provide the highest resolution to date for genetic variation of human populations. This new technology offers great promise for interrogating the genetic etiology of complex disease. However, with this promise come challenges. These new sequencing methods are prone to nontrivial error rates and sparse coverage of mapped reads, confounding polymorphism discovery and genotyping. Copy number variation must often be inferred indirectly. The massive size of these data sets requires rapid and scaleable analytic approaches. In this proposal, we present statistical methods to address these challenges directly, using computationally tractable models for population genetic variation. Our methods take account of the dependence among nearby alleles (linkage disequilibrium) with a clusterbased model for haplotype variation, and utilize this information to aid inferences about the underlying genetic architecture of the samples. Specifically, we propose to call genotypes and detect novel polymorphic loci from next- generation shotgun sequence data, detect rare disease risk alleles for follow-up sequencing studies, and simultaneously model single nucleotide and copy number polymorphism in population data to facilitate studies of association between phenotype and genotype. Our experienced team of medical and statistical geneticists have the technical expertise and access to data sets necessary for achieving these aims. We will implement our methods in our widely-used software package fastPHASE. PUBLIC HEALTH RELEVANCE: High throughput DNA sequencing technology is providing unparalleled detail of human genetic variation. This will allow finer resolution in locating disease genes that affect human health and disease. Both the large quantity and the uneven quality of this new technology demand new statistical methods for inference, risk assessment and eventually clinical translation.

DESCRIPTION (provided by applicant): Pancreatic adenocarcinoma (PaCa) is the 4th leading cause of cancer death in the United States and 8th leading cause worldwide. However, if caught at an early stage, there exist effective surgical treatments. A major difficulty with this is the lck established prevention and screening strategies. Here we propose to discover important genetic risk factors to aid in such a strategy, using exome and targeted sequencing of 4,400 pancreatic cases and 4,400 matched controls of European descent. To manage the cost of sequencing such large portions of the genome, we employ the following 2-stage study design, encompassing our first two aims: (1) deep discovery across whole exomes, followed by (2) targeted sequencing of genes deemed most promising in the first stage. This design retains high power (>90%) to identify genes with moderate risk variation for PaCa, based on the patterns of variation discovered in real studies of breast cancer. In our third aim (3) we propose to quantify somatic mutational load in genes identified from large studies of PaCa genomes, using existing tumor tissue. This somatic variation will be paired to whole-exomes sequenced in Aim 1 to elucidate host-tumor genomic interactions. Our proposed work leverages a strong environment of sample resources at MD Anderson Cancer Center and a well-constructed team of diverse expertise spanning fields of Epidemiology, Genomics, Pathology, Surgery and Computational Human Genetics. The analytical methods we propose will be conducted by leading experts in statistical genetics, who have made major contributions to the development of these techniques. Based on previous studies of familial aggregation of PaCa, and the relative paucity of findings to date from genome-wide association studies, we expect there to be numerous genes with rare variation of intermediate to high risk for PaCa. Given the epidemiological and demographic data, our 2-stage study design and large patient resource, our study is well powered for successful identification of these genes. These results will offer new insights in the etiology of this dreaded and deadly disease.

PROJECT SUMMARY/ABSTRACT The newly-formed Risk, Detection and Outcomes (RDO) Program has 49 members (47 primary, 2 associate) from 19 departments and is led by Drs. Paul Scheet (human genetics, computational biology) with co-leaders Sanjay Shete (biostatistics, genetic epidemiology, population health), Samir Hanash (early detection, proteomics), and Sharon Giordano (health care delivery, outcomes). The major scientific goal of the RDO Program is to reduce the cancer burden in the population and improve quality of life in survivors through innovative research aimed at optimizing cancer risk assessment, screening, early detection, and treatment- associated outcomes from diagnosis through survivorship, with an ultimate goal of informing successful interventions (e.g., in the Cancer Prevention Program). To achieve this goal, the RDO Program is organized into 3 specific aims focusing on 1) Cancer Etiology, 2) Early Detection, and 3) Care Delivery and Outcomes. Aim 1: To discover genetic, behavioral, and environmental factors for cancer initiation. Aim 2: To perform biomarker discovery for personalized risk assessment and early detection. Aim 3: To identify biological and social factors influencing care delivery and patient outcomes. The annual direct peer-reviewed funding of the RDO Program totals $11.4M, including 4 U01s, of which $5.4M (47%) is from the NCI. Over the past 6 years, program members have authored 1211 published peer-reviewed papers, with 373 (31%) intra-programmatic, 594 (49%) inter- programmatic, and 877 (72%) external collaborations. Forty-five percent of articles appeared in journals with IF >5, and 14% appeared in journals with IF >10, including Nat Biotechnol, Nat Genet and J Clin Oncol. Program members used all 14 shared resources. Over this period, the RDO Program has had several major accomplishments. First, in whole-genome genetic epidemiology studies, we identified genetic variants that predispose to disease initiation, affect outcomes, or predict adverse responses to therapy. In multiple whole- exome next-generation sequencing studies, the first of their kind, we are powered to discover variants of higher, intermediate cancer risk. We have also surveyed genomic changes in precancerous tissues, shedding light on early disease pathology. Second, we have uncovered novel blood-based biomarkers for early detection through state-of-the-art profiling technologies. Key hits identified from proteomics and metabolomics promise to complement low-dose CT scans in individuals at high risk for lung cancer. Third, as leaders in a consortium of Texas academic institutions and the Texas Cancer Registry, we have studied patterns of screening, diagnosis, treatment (e.g., chemotherapy and associated decision-making), and follow-up to impact state-wide policy. For all of these endeavors, we continue to develop and enhance unique cohorts, including Cancer Patients and Survivors, Mexican American, Premalignant Genome Atlas, Childhood Cancer Survivors, and organ-specific cohorts of the lung, breast, and ovary.