* simurg/core/brand.py: single source of the mark geometry, colours, favicon
and the report header band (scripts/brand now load it from there).
* HTML reports (QC report, compare dashboard, methylation QC): offline SVG
favicon, indigo header band with the mark, version and a sha256 of the data
shown; Noto Sans font stack; methylation plots in Okabe-Ito colours.
* README banner, guide logo, brand/blog links; branding brief updated to the
final identity; Noto (OFL) and Okabe-Ito entries in THIRD_PARTY_NOTICES;
CHANGELOG (Unreleased).
Unit tests + P5 QC oracle: 285 passed; ruff and mypy clean.
Frontal dotted simurg on a bead grid: gold body and wings, three-dot crown,
red / gold / green channel plumes with peacock eyes. Detailed (31x27), simple
(11x10) and micro (7x7) drawings; dark, light, gold, black, white versions;
outlined Noto Sans wordmark; favicon and app icons; README banner, social
preview, avatar, featured-image template, report corner mark, CLI banner.
Checks: WCAG contrast table, protan/deutan simulation, icon sizes, family
sheet. Blog featured images regenerated with the new mark for consistency.
Geometry shared in scripts/brand/simurg_mark.py.
Dotted simurg on a bead grid in the channel colors (red/green, gold bird,
indigo), one motif per post; generated by scripts/blog/make_featured_images.py.
User guide: ArrayTrain section, v2 commands in the table, v2 accuracy rows,
updated limitations. README status for v2.0.0. Version 2.0.0 (Python package);
the Rust kernels are unchanged and stay simurg_core 1.1.0. uv.lock committed
(uv lock --check clean).
Full suite on the Pi, file by file under the memory guard: 29 files exit 0
(unit: 282 passed), ruff and mypy clean.
The INFO/extra-FORMAT comparison held ~700 000 gtc2vcf records as text plus two
dicts each (> 4.8 GB, stopped by the Pi memory guard). Now written to a temp
file and kept as compact tuples: peak 3.0 GB; all 5 tests pass.
360 arrays vs NYGC 30x SV: all samples OK (K-027 gate fix), PASS precision
DEL 44 % / DUP 57 % (first run 39 % / 34 % with filters ignored), 61 % without
the one excess_events array; rare DEL sensitivity 30 %; LOH replicated 85 %.
m7g.4xlarge spot, 17 min. docs/validation/V2_cn.md, K-027.
Fixed θ windows mislabelled compressed/asymmetric loci (2-cluster loci: 1418 vs
186 errors). Now 1/2/3-component mixtures chosen by BIC, labelled by HWE
goodness of fit plus weak position and gap priors, split clusters merged.
GSE165845 chr22 vs 1000G 30x (240 arrays): concordance 99.44 % (v1 99.27 %,
vendor EGT 99.46 %), NRD 1.30 % (1.89 %, 1.23 %), call rate 99.61 %.
docs/validation/V2_arraytrain.md, K-030; plan §12, CLI-038.
- scripts/pgx/wgs_star_truth.py: star alleles from the 1000G 30x phased panel with the DB definitions
(wildcard-aware haplotype assignment; checked against GeT-RM on the demo CEPH individuals).
- validate_star_getrm.py: --all-genes, --individuals, --coverage (projection onto what the array can see);
replicate_concordance.py.
- 120 individuals x 3 GSA v1 arrays: array-resolvable 1637/1710 (95.7 %), top 1223/1800 (chip lacks many
defining variants), GeT-RM subset 28/35, replicates 99.2 % without the failed array (V2_pgx_star.md, K-029).
noise-outlier exclusion + robustness test (PCN-017/EDGE-026), platform from the CSV manifest (CLI-032)
Also: star-allele calls are checked for one manifest/build per run (PGX-011), duplications at the CN cap are
named xN (PGX-022), and tandem names attach supporting variants to their full copy.
fetch_sources.sh downloads PharmVar per-gene zips; pgx db build adds core alleles for genes CPIC does not
define (normalised against the FASTA, *1 = reference). Masks re-derived with the new variants. GeT-RM:
CYP3A4 15/15, CYP2A6 7/9; all comparable genes 154/168 (CYP1A2/CYP2C8 definitions differ from GeT-RM's).
Sub-region CNs (p5, intron 2, exon 9) resolve full copies and *36/*68/*13 units; candidates carry the
tandem names, background alleles, and CPIC activity sums. GeT-RM: CYP2D6 13/15 top (NA12878 *3/*68+*4 and
NA11832 *1/*68+*4 exact), 15/15 within feasible; all genes 132/144.
- DB build resolves the PGX-007 tag variants from the manifest's probes; coverage over all gene kinds.
- Variant-based genes reported per variant (rs.. REF/ALT | Reference) with the CPIC lowest-activity diplotype;
tagging genes as positive/negative with the population caveat; GSTM1/GSTT1/UGT2B17 *0 from targeted CN.
- Case-insensitive phenotype keys (MT-RNR1). GeT-RM check extended to VKORC1 and the CN genes:
131/144 gene calls (docs/validation/V2_pgx_star.md).
- supporting_variants.tsv, unresolved_variants.tsv, variants_not_covered.tsv with PGX-026 metadata; per-allele
supporting/unresolved sites for both alleles of every candidate; variant_change vs top, distant_alternative,
unexplained_variants, alleles_impacted_by_no_call, alleles_tested.
- simurg pgx coverage: callable / not_covered / masked / equivalent_to per allele and manifest.
- HGVS on GRCh38 RefSeq accessions (NC_...:g.).
CPIC lookups by allele function counts or activity values (duplications N x activity, *5/*0 = 0, total-score
fallback), Indeterminate when feasible solutions disagree or an allele is not in the guideline; per-sample
simurg.pgx/1 JSON and diplotype_annotations.tsv with PGX-026 metadata lines.
- scripts/pgx/fetch_sources.sh: CPIC API tables (CC0) into a dated folder with version and hashes.
- simurg/pgx/db.py: CPIC HGVS locations and values -> left-normalised VCF variants (SNV, del, ins, dup,
repeats, delins, chrM); IUPAC values are core-allele wildcards; alleles without definitions are flagged.
- pipeline/pgx_db.py: deterministic SQLite (content-hash name), probe_map via the SNV VCF site mapper;
GSA v4 PGx: 22 genes, 1354 alleles, 5002 probe links; two builds byte-identical.
The mosaic BAF split is now estimated over heterozygous calls against each probe's expected noise (the
reference heterozygous spread scaled to the sample), inverting the median of |N(+-d, sigma)|. A mosaic loss
lacking the heterozygous calls its PFB predicts gets MF 1 and is promoted (CYTO-020).
Mosaic fraction from the BAF split now subtracts the heterozygous BAF noise (median |BAF-0.5| of unsplit
hets = 0.6745 sigma_b), and a mosaic loss lacking its expected heterozygotes gets MF 1 and is promoted to a
constitutional DEL (CYTO-020). Adds scripts/v2/cn_tune.sh + cn_tune_report.py (GSE165845 rounds on the Pi).
The Pi now boots with cgroup_enable=memory cgroup_memory=1; the MemAvailable guard stays as a
second layer and as the fallback on systems without the memory controller.
scripts/pi_run.sh runs heavy jobs in their own process group and stops them when
MemAvailable drops below a floor (Raspberry Pi OS boots with cgroup_disable=memory, so a
systemd MemoryMax cap is not available); a 360-sample genotype run with cohort QC had
exhausted memory and the OOM killer took down the desktop session.
GSE165845 360 arrays: constitutional DEL/DUP calls 77/74 % in the NYGC SV truth, LOH
reproduced across replicates 74 %; mosaic calls are mostly false positives (12-23 %),
sensitivity for 25 kb-1 Mb events 7-13 % at QUAL >= 20. All samples had been flagged
SampleQuality by the summary logr_dev (cluster-file mismatch); the gate now uses the SD
of the LRR after the batch reference. Validation script: --ignore-filter.
Found by the CN cloud validation: mask_GRCh3{7,8}.bed and cytoBand_GRCh3{7,8}.txt.gz were
never committed nor shipped in the wheel. The ignore rule now covers only the top-level data/.
cn call now records per-chromosome and sample statistics (percent het, LRR stats, BAF
deviations, robust LogR dev, call rate) in the sidecar; cn annotate writes
<sample>.cyto.json (CYTO-033) with ISCN microarray/traditional strings (CYTO-034), bands
and RefSeq gene lists (CYTO-035), LOH summaries, weighted median CN (CYTO-026); annotation
bundle from UCSC cytoBand + ncbiRefSeq (CYTO-036, IN-054). Tests for ISCN strings, gene
overlap, sidecar-missing error and an end-to-end call+annotate.
216 implanted events in 22 synthetic GSA samples (4 replicates per kind x fraction x
size; reference model from 30 other samples): constitutional DEL/DUP 16/16 at 5-20 Mb,
mosaic loss/gain/CN-LOH at 15 % / 5 Mb 12/12.
beta r 0.99998 vs sesame 1.28.1 (max |dbeta| 0.011), detection calls 99.1% identical,
noob within 0.5% median; replicate r 0.997; all 21 control metrics pass. Spec deviations
recorded (5-field control lines, pOOBAH formula, minfi sheet keys). User guide section.
kemal