NovoNordisk - University of Washington Biostatistics

Genetic Analysis Center
Department of Biostatistics, University of Washington
Bruce Weir
Ken Rice
Tim Thornton
Sharon Browning
Brian Browning
Katie Kerr
Adam Szpiro
Cathy Laurie
David Levine
Cecelia Laurie
Sarah Nelson
Stephanie Gogarten
Adrienne Stilp
Caitlin McHugh
Matt Conomos
Quenna Wong
Jean Morrison
Inae Hur
Deepti Jain
Tin Louie
Figure 4
All autosomal SNPs with missing rate < 5% used to calculate PCs
SNP set A
Identity By Descent
• 34 sample pairs with
KC > 1/32
• 10 PO from HapMap
• 21 expected Dups
• 1 unrelated (bottom
• 2 unexpected Dups
Deconvoluting relatedness, population
structure and admixture
Estimate relatedness using KING-robust (robust to population
structure, but not to admixture or departures from HWE)
Partition the sample into a mutually unrelated set and the
remaining (relatives of the unrelated set)
Perform standard PCA on the set of unrelated individuals
Project PC values for the set of related individuals
Re-estimate relatedness using REAP-PC (uses PCs to provide
unbiased kinship coefficients in the presence of population
structure, admixture and HWE departures)
Repeat steps 2-5 to obtain final sets of PCs and kinship
coefficients – to adjust for relatedness and ancestry in
association tests
Matt Conomos and Tim Thornton
GAC Support for HCHS/SOL Genetic Studies
Logistical support for working groups
QA/QC and imputation of genetic data
Estimate relatedness and ancestry variables
Participate in development of paper proposals and
analysis plans by working groups
5. Perform genetic analyses and distribute primary
results to working groups
6. Participate in interpretation of results and manuscript
preparation, which will be led by working group
7. Statistical methodology development
Outline of Genotypic Data QA/QC Process
 Genotyping batch effects (missing call rate, intensity, and allele
 Sample quality (missing call rate, allelic imbalance, heterozygosity)
 Sample identity (gender mismatches, unexpected dups,
unobserved dups, Metabochip mismatches) – resolved 64 of 170
samples with issues
 Chromosomal anomalies (filter those causing genotyping errors)
 Relatedness, admixture and population structure
 SNP quality (missing call rate, Hardy-Weinberg, duplicate
discordance, Mendelian errors)
 Preliminary Association tests (QQ, Manhattan and cluster plots;
expected hits)
 Imputation to 1000 Genomes

similar documents