Updates

v0.3.2

in progress

• Ensures --reference index (.fai) exists
• --fnmax and --pileupmax parameters for exposed special cases
• --minkfreq paramater now reflects ≥ instead of >
• SAMPLE columns cleared on reading to reduce memory usage a little bit (or a lot if there are many samples in input VCF)

v0.3.1

June 25, 2024

• New filtering of haplotypes without paths increases accuracy
• New path scoring improves accuracy and consistency
• ZS and SS FORMAT fields replaced by KS reporting the score
• Requiring reads to span the full variant graph window including --chunksize buffer increases accuracy
• Exhaustive search of partial haplotypes
• Slight runtime reduction from avoidance of redundant path searches

v0.3.0

June 11, 2024

• ~8% speed increase from less work in the path-searching
• Partial haplotypes bug fix increases accuracy
• Fixed SQ and FT fields
• Phase group (PG) field is now canonical phase set (PS)
• Dedicated writing thread helps reduce memory usage by preventing a backlog of completed variants while reading
• Default --out is stdout to allow easier compression/indexing (e.g. kanpig .. | bcftools sort -O z -o out.vcf.gz)
• IUPAC codes are fixed by kanpig according to vcf specifications (Issue #1)
• Fixed filtering of symbolic alts and BNDs
• Argument validation

v0.2.0

May 21, 2024

• Up to 40% reduction in runtime
• Hemizygous and sex chromosome aware genotyping with new --ploidy-bed
• Variants with alternate alleles of stars, monozygotic reference, and BNDs are filtered out
• PathScores now compared with average of size and sequence similarity for increased accuracy
• Cram files now readable

v0.1.2

May 5, 2024

• New optional hompolymer filter doesn't kmerize long homopolymers
• Improved logging info
• Correcting GQ field
• Correcting kmer counting
• Small speed/memory/io improvements
  • Off-loaded annotation work from the single writer thread to the worker threads and using a large multiple of page size for the BufWriter capacity
  • Fewer bam file opens

v0.1.1

Apr 11, 2024

• The --no-prune flag has been changed to --prune since not pruning is a better default.
• Partial haplotypes now only allow up to 3 false negatives for regions with fewer than 500 pileups. More than 500 do not attempt partials.
• Partial haplotypes now respect the --kmer option.

v0.1.0

Apr 9, 2024

Initial version. Works well enough to freeze.