v1.9.0

Added

• --filter-file option on analyze. Accepts a plain-text file with rsIDs and gene names (one entry per line) for custom-panel filtering.
  • Lines matching ^rs\d+$ (case-insensitive) are treated as rsIDs; everything else is a gene name.
  • Comments (#) and blank lines are ignored.
  • Gene and rsID matches combine with OR — an annotation passes if it matches either set.
  • An empty filter file (or one with only comments) produces an empty report. Empty set means "match nothing", distinct from no flag, which means "no filter".
  • Input is case-tolerant; output is canonical (rsIDs lowercased, gene names uppercased) so the recorded filter looks identical regardless of how entries were typed.
  • Not added to methylation or pharmacogenomics, which already operate on curated panels.

Full changelog: https://github.com/dial481/allelix/blob/main/CHANGELOG.md

v1.8.4

Added

• --no-cadd flag on analyze, methylation, and pharmacogenomics commands. Per-invocation CADD enrichment exclusion. Required for commercial deployments without a CADD license from UW CoMotion (CADD is commercial_ok=False). Mirrors the existing --exclude-snpedia licensing-exclusion pattern.

v1.8.3

Changed

• README quickstart. Lead with pip install allelix for end users. Development setup moved to the Development section.
• PyPI publish workflow hardened. Pinned pypa/gh-action-pypi-publish to full commit SHA (v1.14.0) instead of moving branch ref.
• README links. Relative markdown links replaced with absolute GitHub URLs so they resolve on PyPI.
• Security policy scope. Narrowed supported versions to current minor only (1.8.x). v1.x feature line is frozen at v1.8.3.

v1.8.2

Fixed

• HTML report link colors. Links in dark mode were browser-default neon blue (#0000EE), unreadable against the dark background. Dark mode now uses #93c5fd; light mode uses #1976d2 (matching the existing accent).
• Favicon SVG rendering. The report's inline SVG favicon used a linearGradient with an internal IRI reference (url(#g)) that doesn't resolve inside data: URIs, rendering the icon invisible. Replaced with solid-fill landing page SVG design.
• Genotype column header. Abbreviated "Genotype" to "GT" to prevent column header overlap on narrow/mobile screens.

Changed

• PyPI metadata. Homepage now points to allelix.io. Added Source and Changelog links.
• Automated PyPI publishing. GitHub Actions workflow publishes to PyPI via Trusted Publishing on every GitHub Release.

v1.8.1

What's Changed

• Updated sample reports — demo_report.html and demo_report.json regenerated from current databases
• Fixed test fixture genotype format — genotype_match values changed from slash format (A/G) to concatenated format (AG) to match actual annotator output

v1.8.0

What's New

HTML Report Redesign

• 12-column scrollable table → 5-column layout (Magnitude, Gene, Genotype, Repute, Summary)
• Annotations grouped by variant: multiple sources merge into one row
• Clicking a row opens a sliding detail sidebar with all fields displayed vertically
• Variant metrics (frequency, AlphaMissense, CADD) shown once at top of sidebar
• Annotations grouped by source under a single header per source

Dark / Light Mode

• Follows system preference (prefers-color-scheme) by default
• Manual toggle button for override
• All CSS via custom properties

Detail Panel Improvements

• CADD shows PHRED percentile tier inline (e.g. "38.0 (top 0.1% most deleterious)")
• AlphaMissense shows classification alongside score
• CADD/AM thresholds documented in Understanding Magnitude Scores legend
• Repute sort by severity (Bad → Neutral → Good), not alphabetical

PLINK Export Fix

• Variants sorted by chromosome before export — fixes split-chromosome rejection by PLINK1.9

Other

• Inline SVG helix favicon in report
• ClinVar condition "." placeholder normalized to empty
• Reading This Report section collapsed by default
• Embedded JSON uses numeric types and v4-aligned field names
• 9 new JS-HTML contract tests guard against selector/field-name drift
• 1,264 tests passing

v1.7.0

Added

• PLINK export (#29). allelix export plink converts any supported genotype format to PLINK1 binary (.bed/.bim/.fam) for downstream tools (plink2 PCA, ADMIXTURE, PRSice). Single-sample, SNP-major encoding. Uses gnomAD ref/alt for allele coding when available. No-calls and indels skipped.

Fixed

• Multi-allelic strand collision in PLINK coord selection. At sites where one alt is the complement of another (e.g. ref=G, alts=A,T), single-pass coord selection picked the complement match over the forward match — same bug class as CADD #45. Fixed with a two-pass loop that prefers forward allele matches.
• Indel fallback in PLINK export. Multi-character alleles (indels) are now skipped instead of producing non-standard BIM rows.

Documentation

• ADR-0034: Magnitude scoring scale and ceiling (#23). Formalizes the 0-10 scale with practical ceiling of 9. Documents max-across-sources composite rule and reserves per-source scoring for v2.0.0.

v1.6.1

What's new

• Zygosity column — Heterozygous / Homozygous / No Call labels in terminal, HTML, and JSON reports
• CADD PHRED styling — color-coded severity tiers in HTML reports (all four tiers: ≥30, ≥20, ≥10, <10)
• Methylation panel expansion — 27 → 34 genes with cited sources
• Config path display — allelix config set now prints the config file path

Bug fixes

• CADD multi-allelic scoring (#45) — at multi-allelic sites, CADD now looks up the user's specific allele instead of max-reducing across all alts

Schema

• JSON schema version 3 → 4 (new zygosity field; ADR-0033 formalizes bump policy)

Test suite

1,245 tests, 0 failures.

v1.6.0

What's new

CADD v1.7 variant deleteriousness scores (ADR-0032)

PHRED-scaled scores ranking how deleterious any single-nucleotide variant
is, using 100+ annotation tracks (coding, non-coding, regulatory). PHRED
10 = top 10% most deleterious, 20 = top 1%, 30 = top 0.1%.

Two modes:

• Cache (default): pre-built SQLite from HuggingFace (~5 GB, ~120M
  variant keys). Covers all positions in gnomAD, AlphaMissense, and
  ClinVar.
• Full (options.cadd_full = true): queries the complete 81 GB tabix
  file via pysam. GRCh38 only. Adds coverage for novel variants not in
  any pre-computed database.

Opt-in — disabled by default. Enable via allelix db update --cadd or
allelix config set sources.cadd true.

Licensable-source gating for commercial users

Three-state permission model replaces the binary non-commercial gate:

• ALLOW — source is usable.
• BLOCK_PURCHASABLE — blocked in commercial mode, but a license can
  be purchased. config show displays the purchase URL.
• BLOCK_FINAL — blocked with no commercial license available.

CADD: non-commercial by default, unlockable with a UW CoMotion license
(allelix config set license.cadd true). SNPedia: permanently blocked
in commercial mode (no license available).

Other additions

• allelix config get — read a single config key or dump the full config.
• Strand normalization — resolve_strand() maps array-reported
  alleles to reference-forward orientation using gnomAD ref/alt as
  ground truth. Palindromic SNPs return None rather than guessing.
• CADD cache build script — scripts/build_cadd_cache.py filters
  the full CADD SNV and indel files to positions present in gnomAD,
  AlphaMissense, and ClinVar (GRCh38). Uses int64 packing for SNV keys.

Stats

• 1215 tests, 92%+ coverage
• CI green on Python 3.11 + 3.12

v1.5.3

Fixed

• Added missing GWAS Catalog attribution to HTML and JSON reports. GWAS Catalog was registered as an annotator but absent from the license attribution maps in both renderers.

Added

• SECURITY.md. Vulnerability reporting policy (GitHub private vulnerability reporting), supported versions, and scope definition.

Changed

• License metadata centralized on annotator base class via LicenseDescriptor (ADR-0031). Each annotator now declares its license as a required license ClassVar. Non-commercial gating derived from SPDX identifier instead of a hand-maintained set. Report attribution text generated from the descriptor at render time. The NON_COMMERCIAL_SOURCES frozenset in config.py is deleted.
• JSON schema version bumped to 3. The license_attributions block now carries source_url (source website) and license_url (license deed) as separate keys. license field uses SPDX identifiers. Diff between v2 and v3 reports still works.
• CI now gates dev branch pushes (lint + test), not only pull requests.