CHANGELOG.md

Unreleased (2026-04-04)

Bug Fixes

• revert dynamic spectrum-based thresholding to restore variant calling sensitivity
• bug in Graph::BuildGraph: correct AddNodes label

New Features

• Add multi-scale LCR scoring with dual k-mer sizes and calibration tooling
• optimize spoa parameters for local variant extraction

Performance Improvements

• Hoist spoa::AlignmentEngine to VariantBuilder for per-thread reuse
• Eliminate per-read heap allocations via zero-copy bam1_t proxying and uniform paired downsampling
• optimize hot-path allocs & access patterns

Refactoring

• add mike versioning, sync docs, and clean up release tags
• use same prealloc
• only compile gperftools in profile mode
• remove log message info
• Update release notes in changelog
• Update release notes in changelog
• minor fixes to bump helper script
• Update release notes in changelog
• minor fixes to bump helper script
• minor fixes to bump helper script
• use different semver helper tool
• add profile mode to enable cpu profiling in release mode

v2.8.7 (2026-03-30)

Bug Fixes

• use ml model from gcs

Performance Improvements

• use absl::chunked_queue instead of deque

Refactoring

• Bump version to v2.8.7
• update dependencies
• remove pkl from gitattrs
• point to gcs link
• update docs
• add link to ebm model
• add note about cpu support
• remove section in docs
• update to full text link
• remove creation date
• update docs index page
• add biorxiv publication to docs
• add umami scripts block
• update docs with somatic scoring guide
• rename file
• remove redundant f-strings
• add release badge
• update docs system & documentation
• change default arch to skylake-avx512

v2.8.6 (2025-02-06)

Bug Fixes

• remove lowscore filter
• use only stm repo data directory
• Handle case when no reads in window
• handle potential math domain error
• use only class without score
• resolve bug in pysam api usage and add some log msgs
• compute min and max before changing for mad
• resolve typo
• resolve typo in version inc
• update vcf header for clarity
• remove rounding in phred
• make downsampling more predictable
• ensure min chunk length for fuzzy matching
• use 90% of chunk length for fuzzy match cutoff
• count longer indels better with using min of var and read len
• reserve vector length before use
• return untouched ref and qry
• resolve typo in logic
• update vcf header text
• simplify and remove max normal filter
• resolve typo in if statement
• reduce flank to 3 exact match
• remove str specific filter
• check for non gap alignment of flanks instead of match
• reduce AS/XS diff to 1 percent
• reduce AS/XS diff to 5 percent
• reduce AS/XS diff to 10 percent
• increase min difference between AS and XS
• use same value for ends and flank match
• revert back remove short link
• reduce min uniq seq len before trim links
• resolve bug which fails to compile with latest btree
• resolve bug in pre loop value
• increment kmer len by cli step len value
• re-run docs action on change in config
• reduce default min tumor cov filter
• use VCFv4.3 and skip adding MQ0 reads
• resolve undercounting ref for long alleles
• resolve score overflow issue
• resolve typo in while check
• count end gaps and handle it correctly in var ranges
• use more conservative assembly to ref scoring
• create parent path of output vcf if needed
• revert back to v1.14 libdeflate because latest version breaks build
• resolve typo in spdlog usage
• skip regions which have 0 or more than 3 tokens
• include cmake install line
• use debian as default base image
• use relative path to inject right require on build Former-commit-id: 96128bd509124cbd0573dac47a29eacbfe8acb9c Former-commit-id: 40c9b35f7aa984b7b79c6614064cd03ce1aaec5d Former-commit-id: 74cf99ef548050f677de5cce6da1d9ebb9bea6a6
• change config to include static assets Former-commit-id: 468c8c6669da39dcc3a94671db7547a03609a238 Former-commit-id: c65a78edf57ae926af2bee2ee65f13913b8a32a0 Former-commit-id: e8e292b5af811cd57aa7e2f4536244f10113ab1d
• use public api to refer to transcipt member variable Former-commit-id: c0a546e87d8b7b105b4a151ca36e001c6b21f0d0 Former-commit-id: d51fb61953e975c2ed71a5f9c3d031346e6bed41 Former-commit-id: 9cc4265b74796c1f7f3eae08e397890ed611b518
• use node fill colors consistent with v1 Former-commit-id: d6e674c494ec0f1002fd64c3b2c96696c23354f6 Former-commit-id: 1691140cbaf41b10c3649af78fa0b458945c0d3b Former-commit-id: 073c09532824d6984acea58824197e7ecc65216d
• alternate index file missing file extension dot Former-commit-id: dbe5d677498947db73be612a514fffcd7a7ff464 Former-commit-id: 248c5500dbd3141600d610f207cf8099f30f112d Former-commit-id: 240652b1a022c224731387819a7ff0d58a95ee48
• use separate kmer sizes for ref and alt
• resolve typo which removes all normal reads
• use uniq kmer for path seq
• use bq pass only for low quality singleton normal alt snvs
• check if all low qual bases in normal, skip left/right flank for indels
• resolve logic for when no flank is needede
• resolve logic error in if block
• use bq filter for normal only when one low qual SNV alt base found
• use bq filter for snvs and tumor reads
• use bq filter only for snvs
• use bq filter for all single allele haplotypes
• 0 based ref window excludes end, so remove +1
• use right ref and qry indexes after trimming alignment (#3)
• only check high qual snv bases in tumor
• skip adding to kmer coverage for snvs in low qual bases
• add header lines for base qual filters
• remove redundant param
• remove overlap check for read pairs
• always use overlapping reads to increase recall, skip adding empty readInfos
• use right binary path
• go back to full left flank to undo regression in perf
• use small fixed left flank to improve recall
• remove isSomatic ? 0 blocks to fix logic
• use index (0-based) instead of pos (1-based) to pick the spanning node
• break when best path is confirmed
• dont skip nodes in component in remove tips
• swap data with temp to force release of hash table memory
• use vector and .at() to remove undefined behavior
• use padded region end to build windows
• handle unmapped alignments where read and/or mate is unmapped
• use flush ref window instead of index
• handle all window exceptions to prevent deadlock in main thread
• skip on invalid iterator before checking for node ptr
• adjust anchor start on trimming, ensure non bases are not added
• skip early to prevent operating on invalid nodes
• skip returning erased/rehashed node ptrs
• increment index to flush even when no variants are flushed
• allow for merged data source/sink in checks
• trim end gaps in alignments if they exist
• use right positons to point to mismatches (0-based aln start)
• remove -march=native to allow cross compatability
• remove alpine and use ubuntu/debian based builds to fix SIGILLs
• add mutex to log calls to prevent interleaved messages
• move log message before removing
• return from microassembler thread on error
• use shared_ptr without going out of scope
• use right syntax to get value
• use right syntax to get value
• impose strict weak ordering in sort
• add option to skip truncated reference windows

New Features

• use 10 and 13 cutoffs & always write non pass variants
• add script to prep all necessary files for STM viz
• bundle ml model with repo
• bundle ml model with repo
• update filter script to use 10_15 cutoffs
• bundle ml model with repo
• bundle ml model with repo
• increase max window coverage per sample to 1000
• update python deps and filter_variants script to use latest model
• use only hq reads to genotype variants
• update ml model features in apply script
• set qual >=5 i.e. ~70% probability of somatic as cutoff
• only output PASS variants to stdout vcf
• add script to apply somatic ML filtering model
• expand str info into flag, unit length and motif
• expand stats by adding min, max and median abs. deviation
• remove vcf filters & post call filtering to incorporate ML filtering
• use low qual reads for GTing & vcf stats but not for graph
• use same read filters for tumor and normal
• Add window level FMT fields and convert Int to Float
• add fuzzy matching for genotyping long indels
• genotype insertions where read is not fully contained
• change default kmer min, max and step
• add indel specific fisher score threshold
• add strict normal vaf filter for STRs
• double min odds ratio for str variants
• use 1% nml vaf threshold in STR
• add AS/XS filter with 10% difference threshold
• add flank match check to improve precision
• add back short link removal with very lenient params
• add back AS-XS, XT/XA tag read filters
• add kmer step size cli flag
• add rname hash to prevent double counting
• add median
• Fix scoring to call variants longer than read length
• also check for non repeat kmer in refseq
• do some basic read filtering on normal
• add left and right flank haplotypes for coverage count
• require atleast 25% overlap to include read in window
• use avg base qual for both snvs and indels
• reduce min tumor vaf callable
• add low avg base qual filter and info fields
• add contig header lines to vcf output
• write bgzf compressed vcf and auto index at end
• use x86 and arm specific pause/yield instructions
• add flag to use only fully contained reads
• skip adding reads not overlapping region
• use weighted harmonic mean to merge node counts
• write per window contigs fasta
• write one path flow dot file with different hues
• add option for cpu profiling
• use spinlock instead of mutex & use relaxed lock when blocking is not needed
• add option to use overlapping reads to build graph
• catch and rethrow thread exceptions to show failures
• use spdlog instead of custom logger
• use concurrent queue for better window result and progress
• add spython converted singularity recipes
• use concurrent queue for pull-based worker model

Performance Improvements

• optimize graph build by removing cord, turn off secure mimalloc
• optimize queue operations
• throttle counter and reduce default window & overlap
• batch variants/results to reduce lock contention
• use bool array to reduce allocs
• use inlined vector to reduce allocs for small path builders
• add consumer token when dequeueing windows
• use shared_ptrs to reduce window copies

Refactoring

• Bump version to v2.8.6
• update dependencies
• add somtic model and filter script to docker image
• add somtic model and filter script to docker image
• minor fix to prep_stm_viz.sh
• temp remove for lfs tracking
• temp clean git index
• update ide config files
• update dependencies
• update default version string
• update ide files for new dev vm
• Update Dockerfile
• resolve multiple clang-tidy warnings
• resolve multiple clang-tidy warnings
• Bump version to v2.8.5
• update cmake version and use gcc14 for docker build
• update dependencies
• update Dockerfile
• update dependencies
• add pyproject toml details
• Bump version to v2.8.4
• temporarily remove score cutoff
• update conda env deps
• Bump version to v2.8.3
• allow logging htslib errors
• make minor imports
• update cmake version in dockerfile
• update deps & remove extra hash deps
• Bump version to v2.8.2
• update dockerfile
• update dependencies and remove v prefix
• Bump version to v2.8.1
• Bump version to v2.8.0
• Bump version to v2.7.0
• update deps
• Bump version to v2.6.0
• remove extra commands in version extract
• Bump version to v2.5.0
• ensure we don't go over max kmer len
• update deps
• Bump version to v2.4.0
• rename function
• set ends non gap len to 11
• revert back run config
• update deps
• rename actions config
• remove and re-add yml to debug
• Update github-action-deploy.yml
• add styles.module.css
• Add back index.module.css
• Add docker info to README
• auto-use git tag version
• Bump version to v2.3.0
• Bump version to v2.2.0
• remove redundant flag from run config
• remove runtime stats option
• comment out profiling code from release build
• update deps
• use 10mb non gap window for testing
• update dependencies
• update dependencies and docker file
• update deps and dockerfile
• add explicit preset option in genotyper, defaulting to sr
• use max per sample coverage of 500
• rename image name
• update cmake project version
• change cloud build project
• remove redundant cmake configure
• always use commit group type
• Bump version to v2.1.0
• fix typo in dockerfile
• add dockerfile fi
• update dependencies and build tasks
• update local build settings & gitignore
• update CMP0135 policy
• update CMP0135 policy
• update deps
• keep only gcc build Former-commit-id: facbfe77ae435592f8cd0e029e00b8349a2ff446 Former-commit-id: 6d01d6a27eb1c7fdf65dfe33bdbae974bf4d810b Former-commit-id: 4dcb3a240bde0daf58d98d139bd0fb3c9263fe7e
• update graphviz image
• remove ghcr from build to prevent push 403 errors Former-commit-id: 4f83100db7dcea3e5f953573c1ce498831bdc3ed Former-commit-id: a8fd94a3e496a1bc24fd319466b4501e054a38fe Former-commit-id: fae57a5d155e73bc566face970b44d53e6f0cefa
• update readme Former-commit-id: a38e811ade0e6d800a06ec705c295db973eb7150 Former-commit-id: f4f01d634fd6c48fc75cfae8be8dbd5b74ff54b0 Former-commit-id: cc929dc6fc954ca834fa14f32a1ff97cad9b0edf
• Update docs Former-commit-id: a464948922765af128e9f62f302aa3ce5271adf1 Former-commit-id: 80f708b507a345305a5c006b4d6f6514e1c8f65a Former-commit-id: a1302cd2c95236f53c7a5162c0a9caea9d52ea37
• Update docs Former-commit-id: 33f97e8b6a9a54a6fb732dd5462788f36640e7af Former-commit-id: ce5d5a2c9756fe8909d914c61d5ec24f2f0b71a3 Former-commit-id: 93ea17c7d3a5d9ba273c582b2752239a88a05dbb
• Update docs Former-commit-id: 3cc3f6490331cffff8c7c9f526218727364bde10 Former-commit-id: e565f6d9889b716cdd122e3435f7ca633541d134 Former-commit-id: 9d0c32e191be8fef0af955acd649b40e5c2348b3
• Update docs Former-commit-id: 6dc05d640cce51eaecf630749372c868cff5a911 Former-commit-id: 5d4a4b1c7645a3b89a758686ba2af59601dfdcd5 Former-commit-id: abd3d761420e5e6d0c004f5ed366c5969e14f901
• Update docs Former-commit-id: 37d33bcf5ff3165e1700e047bce73f8d8b2f3091 Former-commit-id: 33dbeccd68ef8e375f7fa2ae59c2a20c6f964490 Former-commit-id: 9735acbb78ea328068b708be28cd27cc49caa9da
• Update docs Former-commit-id: 9cf6883ed91fc6954319b9eef2369a2c3e703e65 Former-commit-id: 3a1e40f66fd9a3145632940761ae3c43dc866dfb Former-commit-id: 6a4a699221bc73671d7df2a8607cda6e9fd32cdc
• Update docs Former-commit-id: 1c5ca07cc1317fba1cc7afef87f1c07df4489045 Former-commit-id: d47e8ce4a3cbf063556104d481c6438b5e529d29 Former-commit-id: ce29c3f75a97617c61c41edbd33902d1606b7d7c
• Update docs Former-commit-id: b4f36182263f776aad6007d681da1a7596691a53 Former-commit-id: f7dbf5412c23e2d7f45740bb8300a5a1acdf5c38 Former-commit-id: bfa255ddaf79995a1d126cd011697cdf12cf003e
• Update docs Former-commit-id: 6ea9c8203426e6a344a48a9b1074774b58d19599 Former-commit-id: 20fcca4ef3717bfd7c4fb8852e95fe57e91eeeae Former-commit-id: 5446c5fec6669b4d92ed9213a07ba36dd4daf848
• Update docs Former-commit-id: 8a587117fbdf8b5dc91064a68efa5b7cae7ae004 Former-commit-id: 25255c194fe0cd248490ddff72db792a54f779e5 Former-commit-id: 1b66b168849637c3ddd7037773a69581379024dc
• Update docs Former-commit-id: 6b0b2d9effb71d8394dc41f836916af853070766 Former-commit-id: 54e99f29258b47b3cef70fa7d9aeef39ddd54038 Former-commit-id: 5ede0eda990754840c590afcac290ba85a26ccab
• Update docs Former-commit-id: eed98b81155ae60737996e184912d0de6403461b Former-commit-id: 30d5b3e2a4c763040dd3323137f3186286f790c1 Former-commit-id: 26c3254e281e5e1ff71d857420bf3ed0af15db46
• Update docs Former-commit-id: c31b9397df30264ec156691465636bba3e0aef05 Former-commit-id: a02f16c46d0020d6b1ed5ac653c8deedba3d8016 Former-commit-id: f0cf954c65c0ece715a28c54e54b283acb2f02c0
• Update docs Former-commit-id: 061e979f05de97e1c96d705b20b9681613b61027 Former-commit-id: c68d6163558c27a60906615c6a5ff33afcc3b781 Former-commit-id: 3d14e32fcf3e83950d1e5534df9ee3a45d828537
• Update docs Former-commit-id: 942afe8a454aef8a1364093c8364a8e597ad182f Former-commit-id: a1c6cdb20ad154faeaac9cc07ec5bf583dd9ed75 Former-commit-id: 50b0f7c062d0f7306833d460426474e9904ad70b
• update docs Former-commit-id: 4b7c302b80e12dda4bb8a320a834fa868b05ee40 Former-commit-id: 7190b770e35badc7730974caf37c5d764147ff2d Former-commit-id: b23a08bb6d9ace976ac42d4c1d89ce600d0a3eb7
• update docs config Former-commit-id: e19ab764bc32800f9466372ca28a3fbbc06c5abe Former-commit-id: c205583da1f63b476e552eff0bb74747fb54b9c9 Former-commit-id: 9614b69eac2c6b1f9a1694cd651c308601ee7116
• use permalinks Former-commit-id: c1e9356207ef3f2e95e21fea5a9fc8df11baf9a5 Former-commit-id: fc3ffc309b50c02fa62c01ab61825da09eb2841f Former-commit-id: ae0b1c6e0a42011f10f92deaa8750142472f2fcd
• update visualization section in docs Former-commit-id: 0dda6c83b86b3c80a494de76df1f164a7d76ba5c Former-commit-id: d97e23e1f3e4e8ef7d493577cdebf959d1dff559 Former-commit-id: 866a20e845164ff7cbcb0a6564cf1ba134769e81
• use 0 component for raw graph Former-commit-id: e055d13eb84efec2c8d5693b1f0c43b3ad8dec2a Former-commit-id: c8c60bb1bdc3fd95b610cdfac6dc08680935a6a8 Former-commit-id: 5c8977e2ba9dc06547219b7469213f9084f53172
• also build dot file before low cov removal Former-commit-id: f1f1e775e5949f46e2ac7c80c3f19d6c42017303 Former-commit-id: ab3449a1a319a9f2db05452f14abff3ddf833079 Former-commit-id: 490be6ee4bafe791ad62233b2fd348ad0cf9d953
• ensure non ref alt allele Former-commit-id: e68f2bb6e9c38b1c7926a9781f17ee7988e48c79 Former-commit-id: bdcba126ca1593b7fdcd9673a9ceffe83cac3558 Former-commit-id: 62098deb02f19bae5a4488a181d4b3a7b9df6b26
• use unique path kmer
• use bioinformatics link Former-commit-id: 1fc7c9831a78323c1f348f1a263432a43389e49e Former-commit-id: 2043ca4c05c59bbb57b13ad1dc82b1b53eaf7acc Former-commit-id: 434ccc77009e54405fb7ec14c04d0053dd91f317
• skip low qual normal snvs
• use raw cov for normal always
• change normal read filters
• minor refactoring and edits
• use old merging logic for node labels
• always use bq pass filter for both tmr and nml read snvs
• use overlapping reads
• use contained reads
• make method static
• add new base coverage class
• use non inlined import
• minor edits and refactoring
• add helper canonical hash helper
• replace variants with higher fet score
• remove use contained flag
• remove alt qual from vcf
• reorder loop for perf
• only filter tmr alt quals
• minor edits to fix clang tidy warnings, log trace on alignment erorr
• remove float specifier
• bump to release candidate 1
• use caught exception directly
• add cmake ide profiles to vcs
• update ide files
• add status badge and change workflow name
• fix case for docker image tag
• add default dockerfile and change scope
• add packages write permission
• fix tag names in actions
• push only to ghcr
• run only on release tags and nightly
• build only amd64 images
• remove extra copies from typo
• remove old badges
• update actions yaml to build and push docker images
• remove typo
• update readme
• update dockerfiles to newer compiler versions
• fix clang-12 compilation error
• fix clang-12 compilation error
• fix clang-12 compilation error
• fix clang-12 compilation error
• fix clang-12 compilation error
• add clang tidy fixes
• add license badge and update docs url Former-commit-id: 111a547be0b1e752a58abddb6da0c4fbc97bac60 Former-commit-id: 3fe3a7abd1aded64e727ebbc7155de6cb7a14edf Former-commit-id: a624022b7bd50cfc367251d7d1cf5db9d933b0ec
• update to BSD license Former-commit-id: b5aa128c6537916903d5871051bcfe1be65bee2c Former-commit-id: d810ed6a2f7766c88f0be7455b7688b89dfc7b71 Former-commit-id: 92e0c257074d24c61b5739e25443b942f948f8ff
• add gh pages config
• update dependencies and fix related issues for upgrade
• update ide config files
• ignore variable length check
• always return true when fractionToKeep is 1
• use cflags env var directly
• add local cpp env dockerfile
• add local editor files
• add local editor files
• run configure scripts in bash
• add clion cfg files
• add vscode prefs
• fix diagnostics branch case
• revert concurrentqueue version to fix gcc build error
• temporarily ignore gcc warning
• use newly added absl::StatusOr directly
• use htslib's fisher exact impl
• update deps
• cleanup mark ends and print path flow only if present
• use std::fs and allow existing graphs dir
• minor edits
• update dependencies
• add link to post comparing perf of mutex vs spinlocks
• minor edits and revisions
• add version info on startup
• add --verbose option instead of compile time level selection
• add some context for failed assertion
• add log macros
• minor renaming and edits
• use string vector for regions
• use semicolon separated string as regions input
• revert back to uncompressed outvcf for immediate flushing
• write bgzip compressed output vcf
• flush after roughly every 10000bp
• increase required buffer windows
• skip stripping build files in sanitizer images
• remove build files from docker image, run GH action everyday
• edit file watcher outdir
• fix gcc build errors
• remove LTO and remove deps message on build
• rename queue tokens to clarify usage
• reduce IO operations and reuse file handles per thread
• edit log level for messages
• catch align exceptions
• add LTO for gcc builds
• clarify log message
• add terminate handler to prevent deadlock and add more context in logs
• add log message end of microassembler
• use thread id and single letter level fields in logs
• use env vars to get mimalloc stats
• add another failing complex test case
• fix format string
• remove milliseconds from time format
• output uncompressed vcf and check done bits before breaking
• always skip windows with truncated reference sequence
• tag gcc images as latest
• rehash variant store to ensure consistent memory load
• get vcf prefix and always bgzip output vcf
• allow sanitizer build regardless of build type
• remove unused file from build
• write uncompressed vcf by default
• handle non-ref ends in ProcessPath, skip missing nodes
• check revcomp explicitly and ignore missing buddy neighbours
• continue when neighbour is nullptr
• check in inner remaining loop before merging
• remove alignment gap assert
• skip compressing data source and sink nodes
• add more debug asserts
• continue on errors to debug
• add sanitizer blacklist file
• fix typo and reorder deps
• use alpine linux for release builds after SIGILL fix
• add sha and ref tags only for clang release builds
• include integration test data set
• resolve typo in config
• add thread annotations for tsan
• fix tags for docker images
• run CI/CD only manually
• re-add fixed actions config
• remove config to fix actions trigger issue
• point to right repo name
• add ghub actions config
• cleanup lists and remove unused packages
• change sanitizer build compiler flags and add dockerfiles
• remove lto, remove mimalloc in sanitizer builds
• fix clang builds
• add extra check in debug build
• fix typo in condition
• throw exception to debug state
• flush variants more frequently
• move to debug messages
• use reference order in header or fasta
• add informative log and error message
• log stderr to show feedback without buffer
• initial commit with c++17 codebase

Reverts

• feat: add back AS-XS, XT/XA tag read filters
• Updating website

Pull Requests

• Merge pull request #8 from nygenome/web_docs