Comparison of Variant Calling Pipelines for Whole Genome Sequencing Data in Dairy Cattle

An efficient workflow to accurately call variants from raw sequence data is the first step towards implementing whole genome sequence information in livestock breeding programs. Here we compare currently available variant calling methods applied to a dairy cattle data set. 

Genomic DNA was extracted from semen samples of 66 dairy bulls and sequenced on an Illumina HiSeq2000. Bases were called with the Illumina BaseCaller. The Burrows-Wheeler aligner was used to align paired end reads to the UMD3.1 reference assembly. Duplicate reads were marked using Picard tools. Pipelines integrating data preparation, insertion and deletion realignment, base quality score recalibration and variant calling were set up and compared. Variant calling was done using three different methods: UnifiedGenotyper (Genome Analysis Tool Kit, GATK), SAMTools Mpileup and Platypus. A total of nine different pipelines were compared. Concordance of single-sample and multi-sample calling was analysed between pipelines and by comparison to genotypes from the Illumina BovineHD BeadChip.

Concordance analysis indicated that multi-sample calling performed considerably better than single-sample calling. Insertion and deletion realignment in a “hybrid“ GATK/Mpileup pipeline or the UnifiedGenotyper resulted in higher concordance than that achieved using a stand-alone SAMTools pipeline with considerably more animals. The concordance of variants between all three callers was below 90% indicating that more detailed analysis and/or a combination of callers is necessary. Platypus considerably outperformed both GATK and SAMTools in terms of computation performance.

Financial support from the Swiss Cattle Breeders Association and the Swiss Commission for Technology and Innovation is gratefully acknowledged.