Investigating the Suitability of Some Data Quality Controls Thresholds for Genetic Association Studies of Admixed Population

The purpose of this IRB proposal is to gain access to genetic data generated from participants of publically-funded genomic studies and deposited into dbGaP. It is our intention to use dbGaP data to conduct secondary analysis of the influence of admixture on the outcome of data quality control (QC) in genetic association studies to inform future studies of the optimal QC metric for the genetic association analysis of admixed population. The data we intend to use is deposited in dbGaP and is from the Michigan University Health and Retirement Study (HRS). This protocol is being sent to you because of a dbGaP requirement for this specific de-identified dataset to be reviewed by an IRB as that was not in the protocol.

This work will require the use of statistical analyses tools to estimate the genetic ancestry make-up of each sample, from the genotype data, and determine how that ancestry relates to QC outcomes (i.e. whether or not the sample might be excluded from an analysis due to its ancestral genetic composition rather that the sample genetic quality).

Objectives and specific aims: This work aims to investigate samples failing heterozygosity and sample genotype call rate quality control (QC) to determine whether or not the samples call rate and heterozygosity rate have to do with the ancestry composition of the outliers rather than the quality of the samples and inform future studies of potential loss if general QC is applied to genetic data of admixed sample sets.

Rationale and Background: In genetic association studies DNA sample quality can vary largely across study participants and such variation has an impact on genotype call rate and genotype accuracy; samples of low DNA quality tend to have lower genotype call rate and genotype accuracy. Heterozygosity rate (proportion of heterozygous loci per individual) and genotype failure rate (proportion of missing genotypes per individual) are jointly and routinely used to identify samples with low DNA quality at the data quality control (QC) stage of genetic association studies. Excessive heterozygosity rate may indicate sample contamination whilst a reduced heterozygosity rate could indicate inbreeding [1]. Samples with 3-7% [2,3] genotype call-rate and heterozygosity > 2-3 standard deviations from the mean heterozygosity are usually excluded from genetic case-control studies.

Generally, the sample size of genetic association studies involving minority ethnic groups such as admixed African American tends to be small. It is hence important to ensure samples are not discarded unnecessarily, resulting into reduced statistical power, by using QC thresholds applied to homogenous groups which might not be appropriate for an admixed sample set. The aim of this analysis is to investigate samples failing heterozygosity and sample genotype call rate QC to determine whether or not the samples call rate and heterozygosity rate have to do with the ancestry composition of the outliers rather than the quality of the samples. The motivation is to inform future studies of potential loss if general