basevar v2.6.8

BaseVar v2.6.8 Release Notes

This is a maintenance release focused on code hygiene, a correctness fix in
genotype phase encoding, and a hot-path serialization optimization. VCF output
is byte-for-byte identical to v2.6.7 except for the phase-separator fix
described below.

Bug Fixes

• VCF: encode genotypes as unphased in update_genotypes
• Allele values were always correct; only the / vs | phase separator was
  affected, and only for odd-index alleles (e.g. 0|1 → 0/1, 1|1 → 1/1).

Performance Improvements

• Caller: VCFTextLine::to_string() — eliminate per-line stringstream

Code Hygiene

1. Remove dead to_string() methods

2. Include-what-you-use cleanup

3. Rename caller struct VCFRecord → VCFTextLine

basevar v2.6.7

What's Changed

Build System

• Fix pthread linking when using system htslib (USE_SYSTEM_HTSLIB=ON)
• Add find_package(Threads REQUIRED) for proper threading support

This fixes bioconda build failure on Linux.

basevar v2.6.6

What's Changed

Build System

• Add USE_SYSTEM_HTSLIB CMake option for conda packaging
• Add install() target for proper binary installation
• Add conda/ directory with bioconda recipe files

This release enables bioconda submission.

basevar v2.6.5

BaseVar v2.6.5 Release Notes

Performance Improvements

Caller: BGZF block-level merge for multi-thread and multi-region output

The basevar caller internal VCF merging has been optimized with BGZF block-level copy, eliminating decompression/recompression overhead:

• Thread-level merge (previously merge_file_by_line): per-thread temporary .partN.vcf.gz files are now merged via raw BGZF block copying (bgzf_raw_concat). Since these temporary files contain no VCF header (headerless VCF data streams), the header is written once up-front and the remaining blocks are copied directly.

• Multi-region merge (.vcf.gz output): when the output format is BGZF-compressed, multiple region VCFs are now merged via header-aware block-level copy (bgzf_naive_concat), which writes the header from the first file and skips headers from subsequent files. When the output is plain-text .vcf, the original line-by-line merge is preserved.

Enhancements

concat --naive mode: input/output format validation

• BCF detection: input files starting with non-# bytes (BCF binary format) are rejected early with a clear error message.
• BGZF format check: non-BGZF-compressed inputs are detected and rejected.
• Output suffix validation: output file must have a .gz suffix since --naive mode always produces BGZF-compressed output.

basevar v2.6.4

BaseVar v2.6.4 Release Notes

Bug Fixes

Fix getopt_long optional_argument parsing issue in caller and concat

All options that accept a value in basevar caller and basevar concat have been changed from optional_argument to required_argument in the getopt_long definition.

Affected options (caller): -L, -m, -q, -Q, -B, -t, -r, -G
Affected options (concat): -L

Root cause: With optional_argument, getopt_long requires the value to be attached via = (e.g., --thread=64). Using space-separated syntax (e.g., --thread 64) caused the value to be interpreted as a positional argument, leading to errors such as:

terminate called after throwing an instance of 'std::runtime_error'
  what():  [bam.cpp::Bam:_open] file not found - 64

After fix: Both syntaxes now work correctly:

basevar caller --thread 64 -f ref.fa ...    # space-separated ✓
basevar caller --thread=64 -f ref.fa ...    # equals-sign ✓

All options remain optional — required_argument only means the value must be provided when the option is used, not that the option itself is mandatory.

basevar v2.6.3

New Features

AC_GT / AN_GT / AF_GT — GT-based allele count INFO fields

Added three new INFO fields that count alleles directly from the VCF GT column (posterior genotype calls with population prior):

Field	Type	Description
AC_GT	Integer (Number=A)	Per-ALT allele count from the VCF GT column
AN_GT	Integer (Number=1)	Total alleles in non-missing VCF GT calls
AF_GT	Float (Number=A)	Allele frequency = AC_GT / AN_GT

These complement the existing two tiers:

• Posterior (recommended): AF, AC, AN — expected values from posterior probabilities (with population prior, distinct from GT)
• Observed: AC_obs, AN_obs, AF_obs — discrete counts from argmin(PL) genotypes (without population prior, distinct from GT)

Improvements

VCF header description overhaul

All VCF header descriptions have been systematically audited and rewritten for clarity:

• Removed internal terminology: Replaced "posterior mode", "legacy mode", "hard genotype", and "Bayesian mode" with user-friendly computation-based descriptions.
• GT-source clarification: Added "distinct from GT" to AC/AN/AF and AC_obs/AN_obs/AF_obs descriptions, making it explicit that these are NOT derived from the VCF GT column.
• AF recommendation: The AF field description now includes "(recommended by BaseVar)" to guide users toward the preferred allele frequency metric.
• AN description: Updated to clarify it equals 2 × number of samples.
• AF_group description: Made more generic; removed LRT-specific reference since the field may use different estimation methods.
• BaseQRankSum: Fixed inconsistent capitalization ("Alt Vs. Ref" → "Alt vs. Ref").

VCF 4.2 compliance fixes

• AD: Type=String → Type=Integer (VCF 4.2 spec compliance)
• DP4: Number=A → Number=. (variable-length list, not per-allele)

Single-region VCF output optimization

When calling variants in a single region, BaseVar now directly renames the temp VCF file to the output instead of performing a read-write merge. This eliminates unnecessary I/O for single-region runs.

basevar v2.6.2

BaseVar v2.6.2 Release Notes

Overview

v2.6.2 is a correctness-focused release that fixes two latent bugs in the variant caller's PL (Phred-scaled genotype likelihood) indexing and AC/AN observation counting, and introduces key improvements to the posterior genotype calling mode. All bugs affect multi-allelic sites; bi-allelic behavior is unchanged.

What's New

Dosage-Based AC/AN/AF in Posterior Mode

In posterior mode (--gt-mode posterior, the default), INFO fields AC/AN/AF now report dosage-based values computed from genotype posterior probabilities:

• AF = expected allele frequency from posterior dosage (recommended by BaseVar)
• AC = round(expected allele count from dosage)
• AN = 2 × N_samples

This provides more accurate allele frequency estimates for ultra-low coverage data compared to the previous reads-based counts. Legacy mode (--gt-mode legacy) behavior is unchanged.

Bug Fixes

fix some bugs

basevar v2.6.1

BaseVar v2.6.1 Release Notes

Bug Fix

v2.6.1 is a critical bug fix release that resolves Indel variant loss introduced by the binary batchfile format (v2.6.0). Approximately 2.2% of Indel variants were silently dropped during variant calling due to incorrect ref_base storage semantics in the .bbf binary format.

basevar v2.6.0

BaseVar v2.6.0 Release Notes

Bug Fixes

1. Fix sparse index lockstep position loss
2. Fix VCF coordinate corruption in binary batchfile reader

The fix adds a backfill step after the reader loop to ensure all samples have consistent ref_id/ref_pos values before calling the variant caller.

3. Fix redundant file open in binary batchfile reader

Performance Improvements

1. Binary batchfile read path optimizations

2. Buffer-serialized binary I/O

3. Eliminate redundant seek_virtual

4. Dead code removal

What's New

New basevar dump subcommand

A new dump subcommand for inspecting intermediate binary batchfile (.bbf) and binary index (.bbi) files. This is useful for debugging, verifying file integrity, or examining per-sample read data at specific genomic positions.

# Inspect .bbi index summary (magic, version, entries, position range, footer integrity)
basevar dump sample.bbf.bbi

# List all index entries
basevar dump sample.bbf.bbi --entries

# Show .bbf header (sample IDs) only
basevar dump sample.bbf --header

# Show all positions in compact summary format
basevar dump sample.bbf

# Show per-sample details at a specific position
basevar dump sample.bbf -r chr11:5246595-5246595 -v

# Dump first 20 positions with full per-sample read data
basevar dump sample.bbf -n 20 -v

basevar v2.5.3

BaseVar v2.5.3 Release Notes

Overview

v2.5.3 is a performance optimization release that significantly reduces runtime overhead in the variant calling pipeline through four targeted improvements: eliminating ostringstream conversion bottlenecks, replacing per-base string allocations with char, adding move semantics for BamRecord, and introducing EM hot-start for the step-down procedure.

Upgrade Recommendation

All v2.5.2 users are recommended to upgrade to v2.5.3 for improved runtime performance with no change in output.