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.
Methylation CSV manifest reader (MET-002/003), raw M/U assembly with bead masking
(MET-005), pOOBAH detection p-values, noob background correction, linear/nonlinear dye
bias, beta and M-values (MET-010..012), the 21 methylation control metrics (MET-006,
bs-conv red-bg set option), per-sample ctrl/cgs tables, summary tsv/xlsx/html with the
detection histogram and a beta PCA, pcs.tsv.gz, controls_raw/controls_qc.tsv; methylation
datasets load in qc report (MET-014); qc controls --array-type methylation; minfi-style
Sentrix_ID/Sentrix_Position sheet keys. Vendor EPIC manifests carry a 5th field on some
control lines: >= 4 fields accepted. Runs on the public minfiDataEPIC set (3 x NA12878).
Wheels (pure + manylinux_2_28 kernels), sdist, Docker image and conda packages built and
installed on fresh Ubuntu 24.04 machines; 225 unit tests pass against the installed
package; the tutorial runs with pip (with and without kernels); GTC/VCF content digests are
identical across pip, Docker and conda on both architectures and equal to the Pi's run.
docs/validation/P10.md; guide: install from wheels, Docker, conda; exit code 141.
Found by the P10 fresh-VM run: `simurg inspect … | head` raised BrokenPipeError, reported as
an internal error, and stopped examples/gsa_demo.sh (set -o pipefail). run_guarded now flushes
stdout inside the guard and maps BrokenPipeError to ExitCode.SIGPIPE (141) without a message.
p10.sh: SKIP_DOCKER/SKIP_CONDA, clearer digest report; p10t.sh runs build+pip+tutorial only.
scripts/build_dist.sh: sdist + pure wheel + manylinux_2_28 kernel wheels (maturin --zig,
both architectures). Dockerfile (multi-stage, kernels included, non-root user).
packaging/conda/{simurg,simurg-core}. scripts/p10/digest.py: platform-independent content
digests of GTC/VCF/summary outputs. Cloud task p10 installs the artifacts on a fresh
Ubuntu VM and runs the tutorial with pip, Docker and conda installs. trial.sh picks the
arm64/amd64 image from the instance type and can run instances in parallel (PARALLEL=1).
Licence files ship in both wheels; Rust crate notices added.
Graviton3 c7g.8xlarge, 360 GSE165845 arrays: 360/360 GTCs identical between the Rust
kernels and the pure-Python paths; 1 worker 78-80 samples/min (pure Python 29.6).
Scaling 1->8 workers efficiency 0.87-0.90; above that the 125 MB/s gp3 disk caps the
run (~620 samples/min) and, on a RAM disk, a fixed per-worker start-up cost limits a
360-sample batch to efficiency 0.48 at 32 workers (K-025). Worker state is pickled once.
docs/validation/P9.md, result JSONs, guide/README/CHANGELOG updates.
Rust ports of the hot paths found by profiling: fused per-locus GenCall calling
and scoring, BAF/LRR, libm maps with CPython math semantics, both nearest-neighbour
searches of the normalization fit, and the BPM LocusEntry parser. Address
coverage is memoised per IDAT layout. SIMURG_PURE_PYTHON=1 forces the pure paths.
On the Pi: genotype_sample 3.4 s -> 1.1 s, 36-sample GSA run 131 s -> 46 s, BPM
read 4x faster; 36 GSA + 6 Omni2.5 GTCs and QC tables byte-identical to the pure
path; tests/unit/test_core_kernels.py checks every kernel on edge cases.
bench/profile_genotype.py, bench/scaling.py, cloud task p9 (thread scaling).
docs/user_guide.md (install, inputs, complete run, commands, presets, QC, outputs,
accuracy, limitations, troubleshooting), examples/gsa_demo.sh (end-to-end run on the
public GSA demo, verified on the Pi), examples/sample_sheet.csv, CHANGELOG.md; version 1.0.0.
360 GSA arrays (120 1000 Genomes individuals x 3 replicates): concordance 0.9934 vs
NYGC 30x (0.9958 on sites in both sets), 0.9985 vs GIAB HG001 v4.2.1 (non-ref F1
0.997), replicate concordance 0.9966; allele-swap sites 0.016 %; 358/360 replicate
pairs detected as duplicates. Plan §16.5 metrics in docs/validation/P8.md (K-024).
The download-only phase (EBI ~15 MB/s) left the CPU idle long enough for the
idle alarm to terminate the first run. Streams are now cut to the manifest
positions (±100 bp) and the 120 individuals while genotyping runs.
scripts/p8/prepare.py lays out the 360 GEO IDATs under their header barcode/position;
scripts/p8/validate_p8.py computes concordance, NRD, no-call rate, per-class and per-MAF
sensitivity/specificity/precision/F1 vs NYGC 30x and GIAB HG001, replicate concordance,
systematic/allele-swap site errors, per-chip batch table and duplicate detection.
Cloud userdata can skip the Omni2.5 archive (OMNI=0).
- compare/core: EXACT/NEAR/DISCREPANCY/UNKNOWN with the plan §16.3 tolerances (abs, float32 ULP), NaN = NaN,
locus classes (assay, variant type, chromosome group, intensity-only); fields absent on both sides are
not applicable
- compare/records: GTC vs GTC (per-locus arrays, transforms per bin, controls, sample statistics; metadata
listed, not counted) and VCF vs VCF (records keyed incl. ID; SNV/MNV/indel, merged, multi-allelic classes)
- simurg compare A B (files or folders paired by name): summary.json, discrepancies.tsv.gz (never truncated),
self-contained dashboard.html; --tolerance, --fail-on-discrepancy (exit 3)
- scripts/nightly.py + install_nightly.sh (crontab 02:30): unit tests, GSA vs gtc2vcf (GTC + VCF), VCF
regression vs golden, Omni2.5 subset vs vendor GTCs; first run PASS on every step
- tests/unit/test_compare.py; docs/validation/P7.md, README
- gtc-to-vcf --info-fields ASSAY_TYPE,NORM_BIN,GENTRAIN,CLUSTER_SEP,ILMN_STRAND,PROBE_IDS (--egt) and
--extra-format X,Y,NORMX,NORMY,THETA,R, in per-sample VCFs and the --merge BCF (OUT-012, plan §8.3);
equal to gtc2vcf's values (THETA/R missing where both normalized intensities are 0, K-023)
- --merge-snv-indel (OUT-021, EDGE-013): alleles extended to the longest REF; SNV probes read the base at POS
- --auxiliary-loci (OUT-035, EDGE-015): rules (a)-(e), E_AUX_INVALID; rule (d) uses the last plus-strand base
for minus-strand designs (spec amended, as GTCtoVCF)
- export plink --format bed|pgen|both: PLINK 2 fixed-width .pgen/.pvar/.psam; plink2 reads it back to a .bed
byte-identical to ours
- cohort store keeps raw/normalized intensities for the BCF path; tests (unit + gtc2vcf oracle); docs
- Omni2.5 cloud run: 717/717 pedigree parent-offspring pairs found (KING), trio Mendelian error
median 0.031 %, sex vs pedigree 99.6 %, P4 preset re-check 2138/2141 byte-identical (K-015 closed)
- qc/controls: skip bead addresses absent from the IDAT; control missing only if none is on the chip
- QC-008 amended: signal / zero background = +Inf (passes), 0/0 = NaN (flagged); Inf in TSV/report
- plotly bundle stored gzip-compressed (repo grep stays clean: the DOM event name contains the
forbidden substring)
- docs: validation P5 part 2, summaries, known_unknowns K-021
- vendor comparison (1000 Genomes Omni2.5, AutoCall 1.6.2.2): every
computed field identical in 2138/2141 samples; the 3 others trace back
to the single-locus raw-intensity rule (K-009); sex rule 1 matches all
- AutoCall 1.6.2.x leaves GC10/GC50/#calls/#no-calls at 0 in GTC v3:
new `--preset autocall-1.6.2` (1.1.2 normalization, sex rule 1, GTC v3,
legacy statistics); the sample summary keeps the real statistics
- NaN sign canonicalised to the x86 pattern (K-014); report P4.md
- tests/unit/test_genotype.py: fuzzy memberships incl. NaN rules, calls
at cluster centres, intensity threshold, BAF/LRR interpolation and
NaN pattern, LogR-dev locus set, sex rules v1/v2, TOP-strand base
calls, invariant dates, end-to-end `simurg genotype` (summary schema,
pass-through columns, thread/layout determinism, overwrite refusal)
- tests/oracle/test_p4_genotype.py: 36 GSA samples x 2 versions
byte-identical to the oracle
- scripts/validate_p4_large.py + cloud task p4: compare every vendor
GTC field by field and as whole files; reports which sex rule matches
- genotype/clusters.py: EGT aligned to manifest order (float32 mean R)
- genotype/gencall.py: flair, six-case likelihood, priors/posteriors,
fuzzy sub-scores and score map, vectorised with the reference's
float/double evaluation order and libm transcendentals
- genotype/baf_lrr.py, sex.py (rules v1/v2), sample.py (IDAT -> Gtc)
- pipeline/inputs.py shared by ingest and genotype
- pipeline/genotype.py + CLI: GTC per sample, sample_summary.tsv/json
(OUT-060..063), run manifest
- GTCs from IDATs are byte-identical to the oracle for 36 GSA samples x
2 normalization versions (NaN sign canonicalised to the x86 pattern)
Normalization 1.1.2 reproduces all 109,191 bins of the 1000 Genomes
Omni2.5 GTCs (AutoCall 1.6.2.2) with 0 ULP difference, reserved fields
included; 1.2.0 and 1.1.2 are bit-identical to the no-FMA gtc2vcf oracle
on 36 GSA samples. AWS run: 33 min, ~$0.15, no resources left.
- fit.py: exact and binned nearest-neighbour searches use one distance
matrix per searched bin (first-minimum rule kept); fit 2.6 s -> 1.4 s
per GSA sample, still bit-identical to the oracle
- scripts/validate_p3_large.py: refits 1.2.0 and 1.1.2 transforms for
every vendor GTC and reports ULP distances per field
- cloud trial runs scripts/cloud/tasks/<TASK>.sh (TASK=p3 default;
p1p2 = previous validation + ingest)
- signal/normalization/numerics.py: vendor percentile, trimmean, MAD sigma
(old/new), bisquare robust fits, float32/float64 typing and sequential
sums reproduced exactly (ported from gtc2vcf, MIT)
- signal/normalization/fit.py: outlier removal, offset, rotation, shear,
scale, nn12/rr12, single-channel model for small Infinium I bins,
exact and binned nearest-neighbour sweeps; vendor quirks marked QUIRK
- signal/normalization/apply.py: affine transform and theta/R
- pipeline/normalize.py + CLI `simurg normalize`
- oracle GTCs rebuilt with -ffp-contract=off (aarch64 gcc fuses a*b+c by
default, shifting the last bits); the old FMA oracle differed by <=4 ULP
Result: transforms bit-identical to the oracle for all 36 GSA samples x
48 bins, for both normalization versions (3456/3456 bins, 0 ULP).
Raw X/Y and controls assembled from IDATs equal the vendor GTC arrays for
2138/2141 samples x 2.45M loci; 3 single-locus differences are recorded
as K-009 (probable per-channel missing rule in AutoCall 1.6.2.2). Channel
order X=Red confirmed (U1). `simurg ingest` processed all 2141 samples in
19 min (I/O-bound). AWS run cost ~$0.34 and left no resources.
- tests/oracle/test_p2_raw_xy.py: IDAT-assembled raw X/Y and controls
equal the arrays in oracle GTCs
- validate_p1_large.py: phase-P2 comparison against vendor GTCs with
diagnostics (channel swap, bead-count disagreements)
- cloud trial: also runs `simurg ingest` on all samples (time/memory),
disk 250 GB
All 2141 vendor-written GTC v3 files match BeadArrayFiles field by field
and re-encode byte-identically; all IDAT pairs parse; the 2.45M-locus BPM,
dbv-7 EGT and CSV manifest agree with BeadArrayFiles and each other.
Run on a self-destructing AWS spot instance (42 min, ~$0.20).
kemal