Companion repository for the human variant calling pipelines' benchmark.
Repository contains the following analysis data and scripts:
pipelines folder contains all scripts used to run read aligners and variant callers
eval_happy.sh script was used to run the final benchmarking. eval_happy_42.sh is an analagous script used to evaluate the calls against the newer GIAB v.4.2 reference.
gather_stats.sh was used to gather per-sample hap.py stats into one file.
benchmarking_analysis.R script was used to run all statistical analyses described in the paper.
concordance folder contains scripts used in the analysis of variant call concordance between callers. RTGtools vcfeval was used for all GIAB samples to create a VCF file with TP, FP, and FN calls prior to all downstream analyses.
CustomRegions folder contains stratification BED files that were generated by the authors.
master.bed and master_42.bed files correspond to common regions in GIAB v. 3.3 and GIAB v. 4.2 reference releases.
IMPORTANT! Main files with the statistics generated by hap.py are available through DropBox:
- https://www.dropbox.com/s/jsm6o6dh39oeowj/all_stats_v4_by_strata_rework.tsv.gz - GIAB v. 4.2
- https://www.dropbox.com/s/pr7beov3xzsc4w3/all_stats_by_strata_rework.tsv.gz - GIAB v. 3.3
Other data files in this repository correspond to hybrid selection metrics used to prepare Figure 1 and similar Supplementary figures, variant counts (used in Table 1 and Table 4), results of the concordance analysis (i.e., the final concordance matrix for all GIAB and non-GIAB samples),and coverage model data from Barbitoff et al., 2020.
Citation: Barbitoff Y.A., Abasov R., Tvorogova V.E., Glotov A.S., Predeus A.V. (2022) Systematic benchmark of state-of-the-art variant calling pipelines identifies major factors affecting accuracy of coding sequence variant discovery. BMC Genomics.