The impact of call rate on genotype accuracy

Abstract Text:

Data quality of single nucleotide polymorphism (SNP) arrays plays a key role in the accuracy and precision of downstream data analyses. The use of low quality genotypes can lead to false-positive results and impair the accuracy of genomic predictions. One such quality control measure commonly used is individual animal call rate, defined as the proportion of SNPs per individual where a genotype was called. Currently, no consensus exits on the minimum individual call rate that should be imposed, with threshold call rates per individual varying from 0.80 to 0.95 across studies. The objective of the present study was to determine the minimum individual call rate that could be applied without jeopardising data quality. A total of 144,672 samples genotyped on a custom Illumina genotype panel on 143,827 dairy and beef cattle were available. The genotyping panel includes 14,371 SNPs on either the Illumina Bovine SNP50 or high density genotyping panels. All genotypes were called using the Illumina GenCall method. Lab-dates (n=4) where >15% of the samples genotyped had a call rate <90% were not considered further. Of the remaining 142,433 samples, 493 animals had both a poor call rate (<90%) and a subsequent high call rate (>99%) after re-sampling and re-genotyping. The mean call rate for all samples was 98.77% (range: 15.81%-99.97%). The genotype and allele concordance rate among the genotypes available for all 493 animals with both a low and subsequent high call rate was estimated. Genotype and allelic concordance between low- and high-call genotypes increased as call rate increased (Table 1.). Low minor allele variants (i.e., variants with a minor allele frequency <0.05) were imputed with greatest accuracy for samples with a mean genotype and allelic concordance of 99.24% and 99.55%. Imputation algorithms often correct for genotyping error, therefore to test if imputation improved concordance, all 493 low call rate samples were imputed using FImpute. A reference population of 140,268 animals with call rates >90%, excluding the 493 high-call rate samples, was used. Imputation marginally increased concordance rates for all SNPs with a called genotype, but overall genotype concordance per class slightly decreased because missing genotypes were often incorrectly imputed as heterozygous genotypes (Table 1.). However, if a direct relative (i.e. sire, dam or progeny) was included in the imputation reference population, mean genotype and allele concordance for samples with a call rate between 85-90% increased to 98.13% and 99.04%, respectively.

Keywords: call rate, quality-control, genotype panels