Deep whole-genome sequencing in pedigrees to quantify the contribution of private variants to type 2 diabetes and related metabolic traits. G. Jun1, M. Almeida2, A. Manning3, T. Teslovich1, A. Wood4, M. Zawistowski1, S. Won7, C. Fuchsberger1, S. Feng1, K. Gaulton5, P. Cingolani6, T. Frayling4, G. Abecasis1, J. Blangero2, T2D-GENES Consortium 1) Dept Biostatistics, Univ Michigan, Ann Arbor, Ann Arbor, MI; 2) Texas Biomedical Research Institute, San Antonio, CA; 3) Broad Institute, Boston, MA; 4) University of Exeter, Exeter, UK; 5) University of Oxford, Oxford, UK; 6) McGill University, Montreal, Quebec, Canada; 7) Chungang University, Seoul, South Korea.
Sequencing in pedigrees can be a more effective approach to the analysis of rare and private variants than sequencing unrelated individuals. In T2D-GENES Project 2, we deeply sequenced whole genomes of 590 individuals from 20 large Mexican American pedigrees and imputed the identified sequence variants into additional 448 family members to better understand the role of rare and private variants in type 2 diabetes (T2D) risk and variability in T2D-related traits. Novel (not seen in 1000G or ESP), pedigree-specific singletons in founders are observed on average 2.9 (max. 39) times in the sequenced and imputed data; 16.9% are observed 5 times. To estimate our ability to detect rare variants affecting T2D-related quantitative traits in the given pedigree structure, we simulated singletons in founders that are transmitted into pedigrees, with various effect sizes. For a variant present in a single founder (0.125% frequency), power ranges from 5% when transmitted 5 times up to 43% and 93% when transmitted 10 and 20 times at a level of significance p5x10-8. Assuming 100 variants with a frequency of 0.125% and an effect size of 2 standard deviations (SDs), accounting for 49% of variance, we have 96% power to detect at least one variant, with an expectation of 3.1 variants detected. Our genome-wide analyses identified no variant at p5x10-8 with T2D-related traits, indicating that the number of variants with 2SD effect size would be smaller than 100. When focused on functional candidates and previously identified GWAS regions, we found enriched signals by rare private variants at lower significance levels (10-3p10-5). For example, the strongest associations for fasting glucose are driven by rare variants private to a family across a 13Mb region on chromosome 8, which overlap with a leading signal identified by GWAS (PPP1R3B). We also analyzed 30,567 gene expressions and found 21 transcripts associated with rare variants at p10-10. For 6 transcrips, the most significant association is to a pedigree-specific novel variant with mean effect size 2.6SDs, while the average for all expression transcripts is 0.78SDs. Using gene expression data as proof of principle, we show that deep whole genome sequencing in large pedigrees can identify private and rare associations that would not otherwise be feasible. Results are consistent with a genetic architecture where there are 100 variants with freq. 0.125% that influence diabetes related traits by 2SDs.
You may contact the first author (during and after the meeting) at