milaboratory/mixcr

Major changes

• Major review of affine alignment algorithms

Minor changes

• Fix for MiXCR script for Java 9+
• Fix for exception with no D hits in -dIdentityPercent in exportAlignments (#386)
• Adds warning and fix for performance issue when assemblePartial misused for enriched library

• Chain distribution is back in report file
• V/J chimera detection and elimination mechanism; new parameter minChimeraDetectionScore both alignments in R1 and R2 should be greater than this value
• Smarter second fine alignment step, new parameter alignmentBoundaryTolerance, now this mechanism is triggered even if alignment boundary not perfectly matches read boundary in the opposite mate-pair; minor bug fixes
• Fixed NPE in VDJCHit#getIdentity. Commands like mixcr exportClones ... -cBestIdentityPercent ... were affected
• New align action report line added No target with both V and J alignments; fix for inaccurate value for Alignment failed because of low total score
• Enable asserts for non-release builds, increases chance of detecting subtile error conditions, by trading-off some performance; has no effect on release builds

• Built-in library upgrade to v1.5, many fixes, new species Nannospalax galili (see library release notes for complete list of changes)
• Fix wrong number of chimeric alignments in report with --write-all option
• Removed wrong -v in text warnings in align action
• Added -OmergerParameters.identityType= parameter on align stage (possible values are Unweighted and MinimalQualityWeighted), default value is Unweighted. MinimalQualityWeighted can be used for more reliable merging of paired-end data with low quality ends
• Dropped out-of-the-box support for files produced with repseq.io libraries v1.0 and 1.1

• Fixes possible IOException during report output on Lustre file system

• Fixes bug leading to IOException if report file not exists. (#295)

Major changes

• (beta) Added JSON-formatted reports. Use --json-report option in align, extendAlignments, assemblePartial or assemble actions to create JSON report or append it to already existing report file (jsonl). This feature is intended to simplify batch analysis, and analysis automation. E.g. jq util can be used to aggregate information from multi-json-report files.
• Report file appends are made atomic. Now you can safely specify the same report file name in several invocations of MiXCR, even if you run several instances of MiXCR in parallel. Should also work for NFS and other remote FS, suppporting file locking.
• Added chains column in export (-chains and -topChains)

Minor changes

• Several new counters added to align report
• Added proper filtering for alleles with wildcard symbols in featureToAlign region
• Fixed bug in align report (wrong total count in rare cases) #265
• Documentation added for filterAlignments

• Fixed backward compatibility of *.vdjca files produced by MiXCR 2.1.2 and lower with kaligner2.

• RepSeq.IO updated to 1.2.9, default built-in library updated to 1.4 (read library changelog here and here).

• Minimal thresholds added for V and J alignments in extendAlignments
• minimalNOverlap in assemblePartial changed to 7
• Information on k-mer diversity added to alignPartial report file
• Fix: MiXCR now correctly detects *.fa files as FASTA formatted
• Details of V/J Extension during extendAlignments are now added to target description, and can be retrieved using -targetDescriptions field in exportAlignments