Version 0.15.0

What's Changed

Breaking changes

• The Rust component of STRkit is now required. Pre-built wheels of strkit_rust_ext for some platforms are provided, but otherwise you'll need the Rust toolchain to install STRkit 0.15+.
• Fractional TR genotype calling has been removed.
• Support for specifying more than one alignment file at a time has been removed.

Features and changes

• Caller:
  • Optional incorporation of haplotype-tagged reads from phased alignments
  • Phased blocks of SNVs for tandem repeat phasing
  • Progressive output for JSON, VCFs, and TSVs instead of storing all results in memory
  • Tweaked SNV incorporation logic
  • Added a minimum quality threshold for SNV incorporation
  • Better consensus sequence logic
  • More complete VCF output
  • Call reads with > max reads aligned reads (truncate to max reads reads)
  • Log current processing rate: # loci / second
• Visualization
  • chore(viz): disable search items which don't have calls for now
• Log version on startup

Bug fixes

• Caller:
  • Many important fixes for VCF output
    • fix(call): fix VCF output for alts with no common suffix
    • fix(call): handle bad allele list without crashing in VCF output
  • Fix many issues with reference repeat-counting logic
  • fix(call): properly wait for realign process to terminate on timeout
  • fix(call): better contig format detection for snvs
  • fix(call): bug with terminating progress worker (need to reset stuck count)
  • fix(call): mitigate potential divide-by-zero in locus logging
  • fix(call): using mix of process and main logger in SNV helpers
  • fix(call): properly use process worker, log more info about worker ID
  • fix(call): locking logic + SNV flip logic; make SNV calls tuples
  • fix(call): missing lock call for phase set counter
  • fix(call): missing peak k-mer calculation with SNV/HP peaks
  • fix(call): properly normalize peak weights from GMM medians
  • fix(call): fix peak calling from GMM parameters

Performance

• Caller:
  • Reduce reference sequence re-fetching / reference FASTA accesses
  • CIGAR decoding optimization
  • Repeat-counting optimizations
  • Maximum # of iterations for repeat-counting procedures
  • Approximate repeat counting for very large regions
  • Skip regions which take too long to repeat-count
  • Some memory usage improvements
• Misc. micro optimizations (lambdas, partial functions, etc.)

Documentation

• docs: mention installing as a docker container
• docs: clarify that --realign adds time
• docs: specify that BED file should be sorted
• docs: update alliance install instructions
• docs: mention ONT duplex data
• Update output format documentation

Full Changelog: v0.14.0...v0.15.0