Resequencing Workflows: Variant Calling in Deep and Shallow Coverage NGS Samples

Genotyping-by-sequencing is becoming widespread in breeding research. Efficient data analysis pipelines require that bioinformatics tools are accurate and easy-to-use, and can be optimized and automated.  We describe methodological procedures for calling variants in parental libraries and Recombinant Inbred Lines (RILs) after NGS sequencing.  The parental lines (Lactuca sativa and Lactuca seriola) were sequenced at 12x (deep) coverage, and the 99 distinct RILs at 1x (shallow) coverage. The sequencing was done using HiSeq2000, and the data analysis was performed with CLC Genomics Server and Workbench. The parental libraries were used to create the high quality (HQ) variations track. The reads from L. sativa and L. seriola were mapped separately to L. sativa genomic scaffolds using stringent conditions. The local realignment tool was applied to improve the alignment of individual reads in the presence of indels relative to the reference. The variations were called using the Probabilistic Variant Detection tool (minimum coverage=3, min variant probability=99%). The resulting track/table contained about 15 million variants. To filter for HQ nucleotide variants, additional requirements were applied: control_coverage>2; control_frequency=0; zygosity=homozygous; frequency=100%; coverage<31; type=SNV/MNV. This filtration produced about 11 million HQ variants. For calling variants in shallow sequenced RILs, the same mapping parameters were used, but the stringency of parameters in the variant caller was reduced (minimum coverage=1; variant probability=50%). To select the reliable variants in the RIL libraries, these tracks were compared with the HQ variant track using the “Filter against Known Variants” tool. We compiled these steps into workflows using the graphical workflow creation tool.