After months of brooding over code and results, I’m releasing a new version of the dDocent pipeline (dDocent.FB). dDocent.FB removes almost all BAM processing steps and now uses FreeBayes to call variants, instead of GATK. The change over was largely inspired by an excellent blog post by Brad Chapman. His comparison shows that FreeBayes is comparable to GATK in terms of sensitivity and accuracy, and that going through extra BAM preparation steps like INDEL realignment does not really improve results with haplotype based SNP callers. Removing those extra steps and switching to the faster FreeBayes saves considerable computation time. dDocent.FB is about 10-50 times faster than the previous version!
The repository and User Guide have been updated for this new version. I encourage you to try it out.