This version version greatly lowers Graphtyper's false discovery rate (FDR) and is able to discover more indels than previously. With fewer false variants, genotyping times and memory have been lowered.
Full list of changes:
Graphtyper learned to natively discover from alignments from external aligners using the new "discover" subcommand.
Discovery model has been greatly improved, both lowering overall FDR and increasing indel recall rate.
Graphtyper learned a new subcommand "discovery_vcf" where that can create a disocvery VCF file from serialized variant maps. This removes any batch effects of calling in pools of samples.
Haplotype trimming is now more aggressive, removing any alleles with no confident carriers.
Graphtyper has started learning to genotype SVs (NOTE: experimental at the moment).
Graphtyper learned how to calculate more statistics when --stats option is given, in particular about variant SNP and indel candidates.
The "vcf_break_down" subcommand now makes sure that its output is normalized and sorted. With this change vt is no longer needed.
Graphtyper learned SV genotyping (experimental).
Most of the "vcf_*" subcommands now have the ability to print variants only in certain regions. Begin position is used to deteremine if variant should or should not be printed.
Removed .bis, .ter, etc. endings in variant IDs and use numbers only from now on.
Bugfix: In rare cases the alignment algorithm was not able to align all of the read even if it matched the graph.
Cleaned-up many unused files in the test directory.
This version version greatly lowers Graphtyper's false discovery rate (FDR) and is able to discover more indels than previously. With fewer false variants, genotyping times and memory have been lowered.
Full list of changes: