Variant Calling in Cats and Humans: Are Human-Based Evaluations of Bioinformatic Tools Sufficient?

In humans, the effect of bioinformatic choices on genotype accuracy can be assessed using reference data such as that generated by the Genome in a Bottle consortium. No equivalent tool exists in feline genomes. While many of the fundamental genomic characteristics are similar across mammals, there are differences, and the quality of the reference genome itself differs because of the much larger amount of research into human genes and the human genome. We ask whether the relative accuracy of genotype callers is similar in human and feline genome resequencing. In order to determine this, we look for instances of Mendelian error, in which the genotype of the offspring is not consistent with inheriting alleles from the parents, violating Mendelian principles. In NGS data, Mendelian errors are approximately 1000x more likely to be due to genotyping error than to be the result of de novo mutations, and thus can be used as a proxy for genotyping error, although not all genotyping error causes Mendelian error. Here we have evaluated variant callers including Platypus, GATK Haplotype Caller, and GATK Unified Genotyper.