P9 (WIP): compiled kernels (rust/simurg-core, PyO3) with bit-identical pure-Python fallback
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).
This commit is contained in:
kemal committed
1 parent
f11b39b5a3
commit
6524a6d1b0
19 files changed
+1755
-6
No files matched your search
@@ -23,3 +23,9 @@ jobs:
|
||||
run: uv run mypy
|
||||
- name: Unit tests (no downloaded data needed)
|
||||
run: uv run pytest tests/unit -q
|
||||
- name: Build the compiled kernels (rust/simurg-core)
|
||||
run: |
|
||||
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal
|
||||
PATH="$HOME/.cargo/bin:$PATH" uv pip install ./rust/simurg-core
|
||||
- name: Unit tests with the compiled kernels (must equal the pure-Python results)
|
||||
run: uv run pytest tests/unit -q
|
||||
@@ -0,0 +1,83 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Per-stage timing of `simurg genotype` for one sample (plan §19 measurements).
|
||||
|
||||
python bench/profile_genotype.py BPM EGT GRN_IDAT [--repeat N]
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import datetime
|
||||
import time
|
||||
from dataclasses import replace
|
||||
from pathlib import Path
|
||||
|
||||
import numpy as np
|
||||
|
||||
from simurg.formats.bpm import read_bpm
|
||||
from simurg.formats.egt import read_egt
|
||||
from simurg.formats.gtc import write_gtc
|
||||
from simurg.formats.idat import read_idat
|
||||
from simurg.genotype.baf_lrr import baf_lrr, logr_dev
|
||||
from simurg.genotype.clusters import align_clusters
|
||||
from simurg.genotype.gencall import call_genotypes
|
||||
from simurg.genotype.sample import genotype_sample, manifest_context, params_for_preset
|
||||
from simurg.genotype.sex import estimate_sex
|
||||
from simurg.signal.normalization.apply import normalize, polar
|
||||
from simurg.signal.normalization.fit import fit_transforms
|
||||
from simurg.signal.raw_xy import LocusMapCache, assemble_raw_xy
|
||||
|
||||
|
||||
def main() -> None:
|
||||
ap = argparse.ArgumentParser()
|
||||
ap.add_argument("bpm")
|
||||
ap.add_argument("egt")
|
||||
ap.add_argument("grn")
|
||||
ap.add_argument("--repeat", type=int, default=3)
|
||||
ap.add_argument("--out", default="/tmp/claude-1000/profile.gtc")
|
||||
a = ap.parse_args()
|
||||
m = read_bpm(a.bpm)
|
||||
ctx = manifest_context(m, align_clusters(read_egt(a.egt), m))
|
||||
lite = m.lite()
|
||||
ctx = replace(ctx, manifest=lite, cache=LocusMapCache(lite))
|
||||
params = params_for_preset("current")
|
||||
red_path = a.grn.replace("_Grn", "_Red")
|
||||
t: dict[str, list[float]] = {}
|
||||
|
||||
def lap(name: str, t0: float) -> float:
|
||||
now = time.perf_counter()
|
||||
t.setdefault(name, []).append(now - t0)
|
||||
return now
|
||||
|
||||
for _ in range(a.repeat):
|
||||
t0 = time.perf_counter()
|
||||
grn, red = read_idat(a.grn), read_idat(red_path)
|
||||
t0 = lap("read_idat x2", t0)
|
||||
raw = assemble_raw_xy(grn, red, lite, cache=ctx.cache)
|
||||
t0 = lap("assemble_raw_xy", t0)
|
||||
xf = fit_transforms(raw.raw_x, raw.raw_y, ctx.norm_bins, params.norm_version)
|
||||
t0 = lap("fit_transforms", t0)
|
||||
nx, ny = normalize(raw.raw_x, raw.raw_y, ctx.lookup, xf)
|
||||
th, r = polar(nx, ny)
|
||||
t0 = lap("normalize+polar", t0)
|
||||
z = (raw.raw_x == 0) & (raw.raw_y == 0)
|
||||
thc = np.where(z, np.float32(np.nan), th).astype(np.float32)
|
||||
rc = np.where(z, np.float32(np.nan), r).astype(np.float32)
|
||||
calls = call_genotypes(thc, rc, z, ctx.clusters, lite.intensity_only, params.cutoff)
|
||||
t0 = lap("call_genotypes", t0)
|
||||
estimate_sex(calls.genotypes, r, ctx.clusters.gentrain, ctx.clusters.present, ctx.sex_classes, params.sex)
|
||||
t0 = lap("estimate_sex", t0)
|
||||
_, lrr = baf_lrr(thc, rc, z, ctx.clusters, lite.intensity_only)
|
||||
logr_dev(lrr, ctx.autosomal, lite.intensity_only, ctx.clusters.present)
|
||||
t0 = lap("baf_lrr+logr_dev", t0)
|
||||
g = genotype_sample(grn, red, ctx, params, datetime.datetime(2026, 1, 1))
|
||||
t0 = lap("genotype_sample (whole)", t0)
|
||||
write_gtc(g, Path(a.out))
|
||||
t0 = lap("write_gtc", t0)
|
||||
print(f"{'stage':28s} {'median s':>9s}")
|
||||
for k, v in t.items():
|
||||
print(f"{k:28s} {sorted(v)[len(v) // 2]:9.3f}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,109 @@
|
||||
#!/usr/bin/env python3
|
||||
"""Thread scaling of ``simurg genotype`` (plan §18 P9 gate: ≥ 20 samples/min/core on GSA-size arrays, linear
|
||||
scaling to 32 threads).
|
||||
|
||||
For every worker count the same N samples are genotyped (``--no-cohort-qc``); the per-sample phase is timed from
|
||||
the run log ("genotyping … sample(s)" → "done"), so manifest loading is reported separately and does not blur
|
||||
the scaling. Writes a JSON summary.
|
||||
|
||||
python bench/scaling.py --bpm M.bpm --egt C.egt --idats DIR --out OUT --threads 1,2,4,8 [--samples N]
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import datetime as dt
|
||||
import json
|
||||
import os
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
from pathlib import Path
|
||||
|
||||
|
||||
def phase_seconds(log: Path) -> tuple[float, float]:
|
||||
"""(manifest+setup seconds, per-sample phase seconds) from a genotype run log."""
|
||||
ts: dict[str, dt.datetime] = {}
|
||||
for line in log.read_text().splitlines():
|
||||
rec = json.loads(line)
|
||||
msg = rec.get("msg") or ""
|
||||
t = dt.datetime.fromisoformat(rec["ts"].replace("Z", "+00:00"))
|
||||
ts.setdefault("first", t)
|
||||
if msg.startswith("genotyping "):
|
||||
ts["start"] = t
|
||||
elif msg.startswith("done:"):
|
||||
ts["done"] = t
|
||||
return (ts["start"] - ts["first"]).total_seconds(), (ts["done"] - ts["start"]).total_seconds()
|
||||
|
||||
|
||||
def subset(idats: Path, n: int, dest: Path) -> Path:
|
||||
grn = sorted(p for p in idats.rglob("*_Grn.idat*"))
|
||||
if n <= 0 or n >= len(grn):
|
||||
return idats
|
||||
dest.mkdir(parents=True, exist_ok=True)
|
||||
for g in grn[:n]:
|
||||
d = dest / g.parent.name
|
||||
d.mkdir(exist_ok=True)
|
||||
for p in (g, g.with_name(g.name.replace("_Grn.", "_Red."))):
|
||||
link = d / p.name
|
||||
if not link.exists():
|
||||
link.symlink_to(p.resolve())
|
||||
return dest
|
||||
|
||||
|
||||
def main() -> int:
|
||||
ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
|
||||
ap.add_argument("--bpm", required=True)
|
||||
ap.add_argument("--egt", required=True)
|
||||
ap.add_argument("--idats", required=True, type=Path)
|
||||
ap.add_argument("--out", required=True, type=Path)
|
||||
ap.add_argument("--threads", default="1,2,4")
|
||||
ap.add_argument("--samples", type=int, default=0, help="use the first N samples (0 = all)")
|
||||
ap.add_argument("--simurg", default=str(Path(sys.executable).with_name("simurg")))
|
||||
ap.add_argument("--pure-python", action="store_true", help="run with SIMURG_PURE_PYTHON=1")
|
||||
a = ap.parse_args()
|
||||
a.out.mkdir(parents=True, exist_ok=True)
|
||||
idats = subset(a.idats, a.samples, a.out / "idats")
|
||||
n = len(list(idats.rglob("*_Grn.idat*")))
|
||||
env = {**os.environ, "SIMURG_PURE_PYTHON": "1" if a.pure_python else "0"}
|
||||
runs = []
|
||||
for t in (int(x) for x in a.threads.split(",")):
|
||||
out = a.out / f"t{t}"
|
||||
t0 = time.time()
|
||||
cmd = [a.simurg, "genotype", "--bpm", a.bpm, "--egt", a.egt, "--idats", str(idats), "--out", str(out),
|
||||
"--threads", str(t), "--no-cohort-qc", "--autocall-date", "1/1/2026 12:00:00 AM",
|
||||
"--quiet", "--force"] # fmt: skip
|
||||
rc = subprocess.run(cmd, env=env).returncode
|
||||
wall = time.time() - t0
|
||||
setup, phase = phase_seconds(out / "logs" / "genotype.log")
|
||||
runs.append(
|
||||
{
|
||||
"threads": t,
|
||||
"rc": rc,
|
||||
"wall_s": round(wall, 1),
|
||||
"setup_s": round(setup, 1),
|
||||
"per_sample_phase_s": round(phase, 2),
|
||||
"samples_per_min": round(60 * n / phase, 1) if phase else None,
|
||||
"samples_per_min_per_core": round(60 * n / phase / t, 1) if phase else None,
|
||||
}
|
||||
)
|
||||
print(json.dumps(runs[-1]), flush=True)
|
||||
base = runs[0]["per_sample_phase_s"] * runs[0]["threads"]
|
||||
for r in runs:
|
||||
r["speedup_vs_first"] = round(base / r["per_sample_phase_s"], 2) if r["per_sample_phase_s"] else None
|
||||
r["efficiency"] = round(r["speedup_vs_first"] / r["threads"], 3) if r["speedup_vs_first"] else None
|
||||
from simurg._core import backend
|
||||
|
||||
summary = {
|
||||
"samples": n,
|
||||
"backend": "python" if a.pure_python else backend(),
|
||||
"cpu_count": os.cpu_count(),
|
||||
"machine": os.uname().machine,
|
||||
"runs": runs,
|
||||
}
|
||||
(a.out / "scaling.json").write_text(json.dumps(summary, indent=1))
|
||||
return 0
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
raise SystemExit(main())
|
||||
@@ -0,0 +1 @@
|
||||
target/
|
||||
Generated
+228
@@ -0,0 +1,228 @@
|
||||
# This file is automatically @generated by Cargo.
|
||||
# It is not intended for manual editing.
|
||||
version = 4
|
||||
|
||||
[[package]]
|
||||
name = "autocfg"
|
||||
version = "1.5.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f2032f911046de80f0a198e0901378627c33f59ea0ac00e363d481118bd70a53"
|
||||
|
||||
[[package]]
|
||||
name = "heck"
|
||||
version = "0.5.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "2304e00983f87ffb38b55b444b5e3b60a884b5d30c0fca7d82fe33449bbe55ea"
|
||||
|
||||
[[package]]
|
||||
name = "libc"
|
||||
version = "0.2.189"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "3eaf3ede3fee6db1a4c2ee091bf8a8b4dccdc6d17f656fb07896ee72867612f2"
|
||||
|
||||
[[package]]
|
||||
name = "matrixmultiply"
|
||||
version = "0.3.11"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "3f607c237553f086e7043417a51df26b2eb899d3caff94e6a67592ff992fedc7"
|
||||
dependencies = [
|
||||
"autocfg",
|
||||
"rawpointer",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "ndarray"
|
||||
version = "0.17.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "520080814a7a6b4a6e9070823bb24b4531daac8c4627e08ba5de8c5ef2f2752d"
|
||||
dependencies = [
|
||||
"matrixmultiply",
|
||||
"num-complex",
|
||||
"num-integer",
|
||||
"num-traits",
|
||||
"portable-atomic",
|
||||
"portable-atomic-util",
|
||||
"rawpointer",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "num-complex"
|
||||
version = "0.4.6"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "73f88a1307638156682bada9d7604135552957b7818057dcef22705b4d509495"
|
||||
dependencies = [
|
||||
"num-traits",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "num-integer"
|
||||
version = "0.1.47"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "7ce2d95d4b3734dc35aa2f45e1aa22cd416814592a4f9d9205e11affd5b8e10b"
|
||||
dependencies = [
|
||||
"num-traits",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "num-traits"
|
||||
version = "0.2.19"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "071dfc062690e90b734c0b2273ce72ad0ffa95f0c74596bc250dcfd960262841"
|
||||
dependencies = [
|
||||
"autocfg",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "numpy"
|
||||
version = "0.29.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "6a5b15d63a5ff39e378daed0e1340d3a5964703ea9712eb09a0dc66fade996f4"
|
||||
dependencies = [
|
||||
"libc",
|
||||
"ndarray",
|
||||
"num-complex",
|
||||
"num-integer",
|
||||
"num-traits",
|
||||
"pyo3",
|
||||
"pyo3-build-config",
|
||||
"rustc-hash",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "once_cell"
|
||||
version = "1.21.4"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50"
|
||||
|
||||
[[package]]
|
||||
name = "portable-atomic"
|
||||
version = "1.15.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "05c8b63e8d9609db387f0324918f81d68fe27748f084ef092fb35954d0539a85"
|
||||
|
||||
[[package]]
|
||||
name = "portable-atomic-util"
|
||||
version = "0.2.8"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "10ab3eb7f3becc3a1cbc4f2c6f20267996cfc1a6467a873763411b136a122715"
|
||||
dependencies = [
|
||||
"portable-atomic",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "proc-macro2"
|
||||
version = "1.0.107"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "985e7ec9bb745e6ce6535b544d84d6cd6f7ad8bd711c398938ae983b91a766d9"
|
||||
dependencies = [
|
||||
"unicode-ident",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3"
|
||||
version = "0.29.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4688ddedf473e32662b9b067670129a8afb8c18e351482c70d62ba4a88171e8b"
|
||||
dependencies = [
|
||||
"libc",
|
||||
"once_cell",
|
||||
"portable-atomic",
|
||||
"pyo3-build-config",
|
||||
"pyo3-ffi",
|
||||
"pyo3-macros",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3-build-config"
|
||||
version = "0.29.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f41027e41b4bd03f6e60f9f417fe24a6341a6bb744edd62b6f709f2a52ea30e9"
|
||||
dependencies = [
|
||||
"target-lexicon",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3-ffi"
|
||||
version = "0.29.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "e591a95526fead067432c3b3a33fc74770b87b1e04e73671090d9c2055a2b327"
|
||||
dependencies = [
|
||||
"libc",
|
||||
"pyo3-build-config",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3-macros"
|
||||
version = "0.29.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "73225868fc1cd84eef2c3c230ddb91273bf1de46aeb8a4248da76d32a0924a1c"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"pyo3-macros-backend",
|
||||
"quote",
|
||||
"syn",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "pyo3-macros-backend"
|
||||
version = "0.29.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "571575aa3749fa6216757dd47d2a3e7ef360f329a40f0666a9fbd14889024952"
|
||||
dependencies = [
|
||||
"heck",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "quote"
|
||||
version = "1.0.47"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "1fbf4db142a473a8d80c26bbf18454ed458bf8d26c8219c331daecfdbd079001"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rawpointer"
|
||||
version = "0.2.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "60a357793950651c4ed0f3f52338f53b2f809f32d83a07f72909fa13e4c6c1e3"
|
||||
|
||||
[[package]]
|
||||
name = "rustc-hash"
|
||||
version = "2.1.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "6b1e7f9a428571be2dc5bc0505c13fb6bf936822b894ec87abf8a08a4e51742d"
|
||||
|
||||
[[package]]
|
||||
name = "simurg-core"
|
||||
version = "1.0.0"
|
||||
dependencies = [
|
||||
"numpy",
|
||||
"pyo3",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "syn"
|
||||
version = "2.0.119"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "872831b642d1a07999a962a351ed35b955ea2cfc8f3862091e2a240a84f17297"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"unicode-ident",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "target-lexicon"
|
||||
version = "0.13.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "adb6935a6f5c20170eeceb1a3835a49e12e19d792f6dd344ccc76a985ca5a6ca"
|
||||
|
||||
[[package]]
|
||||
name = "unicode-ident"
|
||||
version = "1.0.26"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "d245f478577f809a851594d02313b640fb437e0bb33866753cff937863096954"
|
||||
@@ -0,0 +1,20 @@
|
||||
[package]
|
||||
name = "simurg-core"
|
||||
version = "1.0.0"
|
||||
edition = "2021"
|
||||
license = "MIT"
|
||||
description = "Compiled kernels for SimurgArray (plan §18 P9)"
|
||||
publish = false
|
||||
|
||||
[lib]
|
||||
name = "simurg_core"
|
||||
crate-type = ["cdylib"]
|
||||
|
||||
[dependencies]
|
||||
numpy = "0.29.0"
|
||||
pyo3 = { version = "0.29", features = ["extension-module", "abi3-py312"] }
|
||||
|
||||
[profile.release]
|
||||
opt-level = 3
|
||||
lto = true
|
||||
codegen-units = 1
|
||||
@@ -0,0 +1,14 @@
|
||||
[build-system]
|
||||
requires = ["maturin>=1.7,<2"]
|
||||
build-backend = "maturin"
|
||||
|
||||
[project]
|
||||
name = "simurg-core"
|
||||
description = "Compiled kernels for SimurgArray (optional; the pure-Python paths give identical results)"
|
||||
requires-python = ">=3.12"
|
||||
license = { text = "MIT" }
|
||||
dependencies = ["numpy>=1.26"]
|
||||
dynamic = ["version"]
|
||||
|
||||
[tool.maturin]
|
||||
features = ["pyo3/extension-module"]
|
||||
@@ -0,0 +1,275 @@
|
||||
//! BPM LocusEntry records — port of `simurg.formats.bpm._parse_locus_entries` (format: Illumina BeadArrayFiles
|
||||
//! `BeadPoolManifest.py`, BSD-2). Same field order, checks and error messages; strings of ≤ 64 bytes are shared
|
||||
//! Python objects, as in the Python parser.
|
||||
|
||||
use std::collections::HashMap;
|
||||
|
||||
use numpy::{IntoPyArray, PyArray1, PyArray2};
|
||||
use pyo3::exceptions::PyValueError;
|
||||
use pyo3::prelude::*;
|
||||
use pyo3::types::{PyList, PyString};
|
||||
|
||||
pub const STR_COLS: usize = 16;
|
||||
// output column order (matches the Python `str_cols` list)
|
||||
const ILMN: usize = 0;
|
||||
const SNP: usize = 1;
|
||||
const CHR: usize = 2;
|
||||
const ISTR: usize = 3;
|
||||
const SSTR: usize = 4;
|
||||
const RSTR: usize = 5;
|
||||
const PLOIDY: usize = 6;
|
||||
const SPECIES: usize = 7;
|
||||
const BUILD: usize = 8;
|
||||
const SOURCE: usize = 9;
|
||||
const SVER: usize = 10;
|
||||
const CSTR: usize = 11;
|
||||
const SSEQ: usize = 12;
|
||||
const TOP: usize = 13;
|
||||
const PA: usize = 14;
|
||||
const PB: usize = 15;
|
||||
|
||||
struct Cursor<'a> {
|
||||
b: &'a [u8],
|
||||
pos: usize,
|
||||
}
|
||||
|
||||
fn err(msg: impl Into<String>) -> PyErr {
|
||||
PyValueError::new_err(msg.into())
|
||||
}
|
||||
|
||||
impl<'a> Cursor<'a> {
|
||||
fn string(&mut self) -> PyResult<&'a [u8]> {
|
||||
let size = self.b.len();
|
||||
let mut p = self.pos;
|
||||
if p >= size {
|
||||
return Err(err("unexpected end of file in locus entry"));
|
||||
}
|
||||
let mut byte = self.b[p];
|
||||
p += 1;
|
||||
let mut ln = (byte & 0x7f) as usize;
|
||||
let mut shift = 7;
|
||||
while byte & 0x80 != 0 {
|
||||
if p >= size || shift > 28 {
|
||||
return Err(err("bad string length in locus entry"));
|
||||
}
|
||||
byte = self.b[p];
|
||||
p += 1;
|
||||
ln |= ((byte & 0x7f) as usize) << shift;
|
||||
shift += 7;
|
||||
}
|
||||
let end = p + ln;
|
||||
if end > size {
|
||||
return Err(err("string runs past end of file in locus entry"));
|
||||
}
|
||||
self.pos = end;
|
||||
Ok(&self.b[p..end])
|
||||
}
|
||||
fn i32(&mut self, what: &str) -> PyResult<i32> {
|
||||
if self.pos + 4 > self.b.len() {
|
||||
return Err(err(format!("unexpected end of file at {what}")));
|
||||
}
|
||||
let v = i32::from_le_bytes(self.b[self.pos..self.pos + 4].try_into().unwrap());
|
||||
self.pos += 4;
|
||||
Ok(v)
|
||||
}
|
||||
}
|
||||
|
||||
type Parsed<'py> = (
|
||||
Vec<Bound<'py, PyList>>,
|
||||
Bound<'py, PyArray1<i64>>,
|
||||
Bound<'py, PyArray1<i32>>,
|
||||
Bound<'py, PyArray1<i32>>,
|
||||
Bound<'py, PyArray1<i16>>,
|
||||
Bound<'py, PyArray1<i8>>,
|
||||
Bound<'py, PyArray1<bool>>,
|
||||
Bound<'py, PyArray1<i8>>,
|
||||
Option<Bound<'py, PyArray2<f32>>>,
|
||||
usize,
|
||||
);
|
||||
|
||||
/// Parse `n` LocusEntry records starting at byte 0 of `buf`. Returns the 16 string columns, map_infos,
|
||||
/// address_a, address_b, assay types (−1 = infer from sequences, entry version < 6), expected clusters,
|
||||
/// intensity-only flags, entry versions, base fractions (None unless some entry is v7+) and the bytes consumed.
|
||||
/// Errors are ValueError with the Python parser's message.
|
||||
#[pyfunction]
|
||||
pub fn bpm_locus_entries<'py>(py: Python<'py>, buf: &[u8], n: usize, names: Vec<String>) -> PyResult<Parsed<'py>> {
|
||||
let name_to_idx: HashMap<&str, usize> = names.iter().enumerate().map(|(i, s)| (s.as_str(), i)).collect();
|
||||
let mut cols: Vec<Vec<Option<Py<PyString>>>> = (0..STR_COLS).map(|_| (0..n).map(|_| None).collect()).collect();
|
||||
let mut intern: HashMap<&[u8], Py<PyString>> = HashMap::new();
|
||||
let mut map_infos = vec![0i64; n];
|
||||
let mut address_a = vec![0i32; n];
|
||||
let mut address_b = vec![0i32; n];
|
||||
let mut assay = vec![0i16; n];
|
||||
let mut exp_clusters = vec![0i8; n];
|
||||
let mut int_only = vec![false; n];
|
||||
let mut versions = vec![0i8; n];
|
||||
let mut fracs = vec![0f32; 4 * n];
|
||||
let mut seen = vec![false; n];
|
||||
let mut any_v7 = false;
|
||||
let mut c = Cursor { b: buf, pos: 0 };
|
||||
|
||||
let to_py = |raw: &'_ [u8], intern: &mut HashMap<&[u8], Py<PyString>>| -> Py<PyString> {
|
||||
if let Some(s) = intern.get(raw) {
|
||||
return s.clone_ref(py);
|
||||
}
|
||||
let s = PyString::new(py, &String::from_utf8_lossy(raw)).unbind();
|
||||
s
|
||||
};
|
||||
|
||||
for _ in 0..n {
|
||||
let ver = c.i32("locus entry")?;
|
||||
if ![4, 6, 7, 8].contains(&ver) {
|
||||
return Err(err(format!("unsupported locus entry version {ver}")));
|
||||
}
|
||||
let ilmn_id = c.string()?;
|
||||
let name = c.string()?;
|
||||
for _ in 0..3 {
|
||||
c.string()?;
|
||||
}
|
||||
c.i32("locus index")?;
|
||||
c.string()?;
|
||||
let ilmn_strand = c.string()?;
|
||||
let snp = c.string()?;
|
||||
let chrom = c.string()?;
|
||||
let ploidy = c.string()?;
|
||||
let species = c.string()?;
|
||||
let map_info = c.string()?;
|
||||
let top_seq = c.string()?;
|
||||
let cust_strand = c.string()?;
|
||||
if c.pos + 8 > buf.len() {
|
||||
return Err(err("unexpected end of file at addresses"));
|
||||
}
|
||||
let a_addr = c.i32("addresses")?;
|
||||
let b_addr = c.i32("addresses")?;
|
||||
let probe_a = c.string()?;
|
||||
let probe_b = c.string()?;
|
||||
let build = c.string()?;
|
||||
let source = c.string()?;
|
||||
let source_ver = c.string()?;
|
||||
let source_strand = c.string()?;
|
||||
let source_seq = c.string()?;
|
||||
let (mut at, mut exp_cl, mut io) = (-1i16, 0i8, 0u8);
|
||||
if ver >= 6 {
|
||||
if c.pos + 4 > buf.len() {
|
||||
return Err(err("unexpected end of file at assay type"));
|
||||
}
|
||||
exp_cl = buf[c.pos + 1] as i8;
|
||||
io = buf[c.pos + 2];
|
||||
at = buf[c.pos + 3] as i16;
|
||||
c.pos += 4;
|
||||
if at > 2 {
|
||||
return Err(err(format!("invalid assay type {at}")));
|
||||
}
|
||||
if (b_addr == 0) != (at == 0) {
|
||||
return Err(err("assay type is inconsistent with address B"));
|
||||
}
|
||||
}
|
||||
let mut f4: Option<[f32; 4]> = None;
|
||||
if ver >= 7 {
|
||||
if c.pos + 16 > buf.len() {
|
||||
return Err(err("unexpected end of file at base fractions"));
|
||||
}
|
||||
let mut f = [0f32; 4];
|
||||
for (k, v) in f.iter_mut().enumerate() {
|
||||
*v = f32::from_le_bytes(buf[c.pos + 4 * k..c.pos + 4 * k + 4].try_into().unwrap());
|
||||
}
|
||||
f4 = Some(f);
|
||||
c.pos += 16;
|
||||
}
|
||||
let ref_strand: &[u8] = if ver >= 8 { c.string()? } else { b"" };
|
||||
|
||||
let name_s = String::from_utf8_lossy(name);
|
||||
let idx = match name_to_idx.get(name_s.as_ref()) {
|
||||
Some(&i) => i,
|
||||
None => return Err(err(format!("locus entry {} not in the names table", py_repr(&name_s)))),
|
||||
};
|
||||
if seen[idx] {
|
||||
return Err(err(format!("locus entry {} appears twice", py_repr(&name_s))));
|
||||
}
|
||||
seen[idx] = true;
|
||||
for (col, raw) in [
|
||||
(ILMN, ilmn_id),
|
||||
(SNP, snp),
|
||||
(CHR, chrom),
|
||||
(ISTR, ilmn_strand),
|
||||
(SSTR, source_strand),
|
||||
(RSTR, ref_strand),
|
||||
(PLOIDY, ploidy),
|
||||
(SPECIES, species),
|
||||
(BUILD, build),
|
||||
(SOURCE, source),
|
||||
(SVER, source_ver),
|
||||
(CSTR, cust_strand),
|
||||
(SSEQ, source_seq),
|
||||
(TOP, top_seq),
|
||||
(PA, probe_a),
|
||||
(PB, probe_b),
|
||||
] {
|
||||
let s = to_py(raw, &mut intern);
|
||||
if raw.len() <= 64 && !intern.contains_key(raw) {
|
||||
intern.insert(raw, s.clone_ref(py));
|
||||
}
|
||||
cols[col][idx] = Some(s);
|
||||
}
|
||||
map_infos[idx] = if map_info.is_empty() {
|
||||
0
|
||||
} else {
|
||||
let t = String::from_utf8_lossy(map_info);
|
||||
match t.trim().parse::<i64>() {
|
||||
Ok(v) => v,
|
||||
Err(_) => {
|
||||
return Err(err(format!(
|
||||
"non-integer MapInfo {} for {}",
|
||||
py_repr(&t),
|
||||
py_repr(&name_s)
|
||||
)))
|
||||
}
|
||||
}
|
||||
};
|
||||
address_a[idx] = a_addr;
|
||||
address_b[idx] = b_addr;
|
||||
assay[idx] = at;
|
||||
exp_clusters[idx] = exp_cl;
|
||||
int_only[idx] = io != 0;
|
||||
versions[idx] = ver as i8;
|
||||
if let Some(f) = f4 {
|
||||
fracs[4 * idx..4 * idx + 4].copy_from_slice(&f);
|
||||
any_v7 = true;
|
||||
}
|
||||
}
|
||||
let missing = seen.iter().filter(|s| !**s).count();
|
||||
if missing > 0 {
|
||||
return Err(err(format!("{missing} loci in the names table have no locus entry")));
|
||||
}
|
||||
let empty = PyString::new(py, "").unbind();
|
||||
let lists = cols
|
||||
.into_iter()
|
||||
.map(|col| PyList::new(py, col.into_iter().map(|s| s.unwrap_or_else(|| empty.clone_ref(py)))))
|
||||
.collect::<PyResult<Vec<_>>>()?;
|
||||
let fr = if any_v7 {
|
||||
Some(numpy::ndarray::Array2::from_shape_vec((n, 4), fracs).map_err(|e| err(e.to_string()))?.into_pyarray(py))
|
||||
} else {
|
||||
None
|
||||
};
|
||||
Ok((
|
||||
lists,
|
||||
map_infos.into_pyarray(py),
|
||||
address_a.into_pyarray(py),
|
||||
address_b.into_pyarray(py),
|
||||
assay.into_pyarray(py),
|
||||
exp_clusters.into_pyarray(py),
|
||||
int_only.into_pyarray(py),
|
||||
versions.into_pyarray(py),
|
||||
fr,
|
||||
c.pos,
|
||||
))
|
||||
}
|
||||
|
||||
/// Python `repr()` of a str for error messages (single quotes unless the text contains one).
|
||||
fn py_repr(s: &str) -> String {
|
||||
if s.contains('\'') && !s.contains('"') {
|
||||
format!("\"{s}\"")
|
||||
} else {
|
||||
format!("'{}'", s.replace('\\', "\\\\").replace('\'', "\\'"))
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,357 @@
|
||||
//! Fused per-locus ports of `simurg.genotype.gencall.call_genotypes` and `simurg.genotype.baf_lrr.baf_lrr`
|
||||
//! (which port gtc2vcf `gen_std_flair`/`modilik`/`compute_score_call_prelim`/`gencall_score_map`/`make_calls`
|
||||
//! and `compute_baf_lrr`, MIT, © Giulio Genovese). Every intermediate keeps the NumPy reference's type (f32 or
|
||||
//! f64) and evaluation order; NumPy's `fmin`/`fmax` are `f32::min`/`max`; libm calls use CPython `math`
|
||||
//! semantics. One pass per locus instead of dozens of full-array temporaries.
|
||||
|
||||
use numpy::{IntoPyArray, PyArray1, PyReadonlyArray1};
|
||||
use pyo3::exceptions::PyValueError;
|
||||
use pyo3::prelude::*;
|
||||
|
||||
use crate::{math_exp, math_log, math_pow};
|
||||
|
||||
const M_2_SQRTPI: f64 = 1.12837916709551257390;
|
||||
const M_SQRT1_2: f64 = 0.707106781186547524401;
|
||||
const NEG_NAN32: f32 = f32::from_bits(0xffc0_0000);
|
||||
|
||||
#[inline]
|
||||
fn smf(x: f32, a: f32, b: f32) -> f32 {
|
||||
let mid = (a + b) / 2.0;
|
||||
if a >= b {
|
||||
return if x >= mid { 1.0 } else { 0.0 };
|
||||
}
|
||||
let u = (x - a) / (b - a);
|
||||
let v = (x - b) / (b - a);
|
||||
if x <= a {
|
||||
0.0
|
||||
} else if a < x && x <= mid {
|
||||
2.0 * (u * u)
|
||||
} else if mid < x && x <= b {
|
||||
1.0 - 2.0 * (v * v)
|
||||
} else {
|
||||
1.0
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn zmf(x: f32, a: f32, b: f32) -> f32 {
|
||||
let mid = (a + b) / 2.0;
|
||||
let u = (x - a) / (b - a);
|
||||
let v = (x - b) / (b - a);
|
||||
if x <= a {
|
||||
1.0
|
||||
} else if a < x && x <= mid {
|
||||
1.0 - 2.0 * (u * u)
|
||||
} else if mid < x && x <= b {
|
||||
2.0 * (v * v)
|
||||
} else {
|
||||
0.0
|
||||
}
|
||||
}
|
||||
|
||||
/// Normal pdf pieces of `_pdf`: (full, peak, bad) for value x, mean mu, sd sigma.
|
||||
#[inline]
|
||||
fn pdf(x: f32, mu: f32, sigma: f32) -> (f64, f64, bool) {
|
||||
let z = ((x - mu) / sigma) as f64;
|
||||
let e = math_exp(-0.5 * (z * z));
|
||||
let sd = sigma as f64;
|
||||
let full = e * 0.5 * M_2_SQRTPI * M_SQRT1_2 / sd;
|
||||
let peak = (0.5 * M_2_SQRTPI * M_SQRT1_2) / sd;
|
||||
(full, peak, sigma <= 0.0)
|
||||
}
|
||||
|
||||
#[derive(Clone, Copy)]
|
||||
enum Side {
|
||||
Left,
|
||||
Right,
|
||||
Full,
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn get(p: (f64, f64, bool), x: f32, mu: f32, side: Side) -> f64 {
|
||||
let (full, peak, bad) = p;
|
||||
if bad {
|
||||
return f64::NAN;
|
||||
}
|
||||
match side {
|
||||
Side::Left => {
|
||||
if x < mu {
|
||||
peak
|
||||
} else {
|
||||
full
|
||||
}
|
||||
}
|
||||
Side::Right => {
|
||||
if x > mu {
|
||||
peak
|
||||
} else {
|
||||
full
|
||||
}
|
||||
}
|
||||
Side::Full => full,
|
||||
}
|
||||
}
|
||||
|
||||
fn check(len: usize, n: usize, what: &str) -> PyResult<()> {
|
||||
if len != n {
|
||||
return Err(PyValueError::new_err(format!("gencall: {what} has length {len}, expected {n}")));
|
||||
}
|
||||
Ok(())
|
||||
}
|
||||
|
||||
/// Genotypes (0 NC, 1 AA, 2 AB, 3 BB) and GenCall scores; (n, 3) cluster arrays are passed flattened (C order).
|
||||
#[pyfunction]
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
pub fn call_genotypes<'py>(
|
||||
py: Python<'py>,
|
||||
theta: PyReadonlyArray1<'py, f32>,
|
||||
r: PyReadonlyArray1<'py, f32>,
|
||||
raw_zero: PyReadonlyArray1<'py, bool>,
|
||||
present: PyReadonlyArray1<'py, bool>,
|
||||
n_cl: PyReadonlyArray1<'py, i32>,
|
||||
theta_mean: PyReadonlyArray1<'py, f32>,
|
||||
theta_dev: PyReadonlyArray1<'py, f32>,
|
||||
r_mean: PyReadonlyArray1<'py, f32>,
|
||||
r_dev: PyReadonlyArray1<'py, f32>,
|
||||
intensity_threshold: PyReadonlyArray1<'py, f32>,
|
||||
gentrain: PyReadonlyArray1<'py, f32>,
|
||||
intensity_only: PyReadonlyArray1<'py, bool>,
|
||||
cutoff: f32,
|
||||
) -> PyResult<(Bound<'py, PyArray1<u8>>, Bound<'py, PyArray1<f32>>)> {
|
||||
let (th, rr, rz, pres) = (theta.as_slice()?, r.as_slice()?, raw_zero.as_slice()?, present.as_slice()?);
|
||||
let (ncl, tmv, tdv, rmv, rdv) =
|
||||
(n_cl.as_slice()?, theta_mean.as_slice()?, theta_dev.as_slice()?, r_mean.as_slice()?, r_dev.as_slice()?);
|
||||
let (ithr, gt, io) = (intensity_threshold.as_slice()?, gentrain.as_slice()?, intensity_only.as_slice()?);
|
||||
let n = th.len();
|
||||
for (l, w) in [(rr.len(), "r"), (rz.len(), "raw_zero"), (pres.len(), "present"), (ithr.len(), "threshold"),
|
||||
(gt.len(), "gentrain"), (io.len(), "intensity_only")] {
|
||||
check(l, n, w)?;
|
||||
}
|
||||
for (l, w) in [(ncl.len(), "n"), (tmv.len(), "theta_mean"), (tdv.len(), "theta_dev"), (rmv.len(), "r_mean"),
|
||||
(rdv.len(), "r_dev")] {
|
||||
check(l, 3 * n, w)?;
|
||||
}
|
||||
let tiny_d = f64::MIN_POSITIVE * f64::EPSILON;
|
||||
let third: f32 = 1.0 / 3.0;
|
||||
let mut geno = vec![0u8; n];
|
||||
let mut scores = vec![0f32; n];
|
||||
|
||||
for i in 0..n {
|
||||
let t = th[i];
|
||||
let r = rr[i];
|
||||
let tm = [tmv[3 * i], tmv[3 * i + 1], tmv[3 * i + 2]];
|
||||
let td = [tdv[3 * i], tdv[3 * i + 1], tdv[3 * i + 2]];
|
||||
let rm = [rmv[3 * i], rmv[3 * i + 1], rmv[3 * i + 2]];
|
||||
let rd = [rdv[3 * i], rdv[3 * i + 1], rdv[3 * i + 2]];
|
||||
let nc = [ncl[3 * i], ncl[3 * i + 1], ncl[3 * i + 2]];
|
||||
|
||||
// gen_std_flair
|
||||
let z1 = 0.5f32 * (tm[1] - tm[0]) / (td[1] + td[0]);
|
||||
let z2 = 0.5f32 * (tm[2] - tm[1]) / (td[2] + td[1]);
|
||||
let mz = z1.min(z2);
|
||||
let alpha = (third * mz).max(1.0);
|
||||
let beta = (third * mz).min(1.0);
|
||||
let eta = alpha * beta;
|
||||
let mut ftd = [0f32; 3];
|
||||
for k in 0..3 {
|
||||
let v = td[k] * eta;
|
||||
ftd[k] = if v < 0.02 { 0.02 } else { v };
|
||||
}
|
||||
let med = td[0].min(td[1]).max(td[0].max(td[1]).min(td[2]));
|
||||
let min_r = if 0.2f32 < med { med } else { 0.2 };
|
||||
let mut frd = [0f32; 3];
|
||||
for k in 0..3 {
|
||||
let mrk0 = rm[k] / 7.0;
|
||||
let mrk = if mrk0 < min_r { min_r } else { mrk0 };
|
||||
frd[k] = if rd[k] < mrk { mrk } else { rd[k] };
|
||||
}
|
||||
|
||||
// modilik
|
||||
let d: [f64; 3] = std::array::from_fn(|k| 100.0 * ((t - tm[k]).abs() as f64) + ((r - rm[k]).abs() as f64));
|
||||
let (daa, dab, dbb) = (d[0], d[1], d[2]);
|
||||
let ok_t = !t.is_nan();
|
||||
let pt: [(f64, f64, bool); 3] = std::array::from_fn(|k| pdf(t, tm[k], ftd[k]));
|
||||
let pr: [(f64, f64, bool); 3] = std::array::from_fn(|k| pdf(r, rm[k], frd[k]));
|
||||
let pdf2 = |k: usize, st: Side, sr: Side| get(pt[k], t, tm[k], st) * get(pr[k], r, rm[k], sr);
|
||||
use Side::{Full, Left, Right};
|
||||
let mut lik = [0f64; 3];
|
||||
let mut covered = false;
|
||||
if ok_t {
|
||||
let cases: [(bool, [(usize, Side, Side); 2]); 6] = [
|
||||
(daa <= dbb && dab <= dbb && rm[0] <= rm[1], [(0, Left, Left), (1, Right, Right)]),
|
||||
(daa <= dab && dbb <= dab && rm[0] <= rm[2], [(0, Left, Left), (2, Right, Right)]),
|
||||
(dab <= daa && dbb <= daa && rm[1] <= rm[2], [(1, Left, Left), (2, Right, Right)]),
|
||||
(daa <= dbb && dab <= dbb && rm[0] > rm[1], [(0, Left, Right), (1, Right, Left)]),
|
||||
(dab <= daa && dbb <= daa && rm[1] > rm[2], [(1, Left, Right), (2, Right, Left)]),
|
||||
(daa <= dab && dbb <= dab && rm[0] > rm[2], [(0, Left, Right), (2, Right, Left)]),
|
||||
];
|
||||
for (cond, assigns) in cases {
|
||||
if cond {
|
||||
for (k, st, sr) in assigns {
|
||||
lik[k] = pdf2(k, st, sr);
|
||||
}
|
||||
covered = true;
|
||||
}
|
||||
}
|
||||
}
|
||||
if !covered {
|
||||
for (k, l) in lik.iter_mut().enumerate() {
|
||||
*l = pdf2(k, Full, Full);
|
||||
}
|
||||
}
|
||||
for l in lik.iter_mut() {
|
||||
if l.is_nan() {
|
||||
*l = 0.0;
|
||||
}
|
||||
}
|
||||
if !ok_t {
|
||||
lik = [f64::NAN; 3];
|
||||
}
|
||||
|
||||
// priors and posteriors
|
||||
let ntot = nc[0] as i64 + nc[1] as i64 + nc[2] as i64;
|
||||
let p: [f64; 3] = std::array::from_fn(|k| (((nc[k] as f32) / (ntot as f32)) as f64 + 1.0) / 4.0);
|
||||
let lp: [f64; 3] = std::array::from_fn(|k| lik[k] * p[k]);
|
||||
let ev = (lp[0] + lp[1]) + lp[2];
|
||||
let (aa, ab, bb) = (lp[0] / ev, lp[1] / ev, lp[2] / ev);
|
||||
let mut iap: u8 = if aa >= ab && aa >= bb {
|
||||
1
|
||||
} else if ab >= aa && ab >= bb {
|
||||
2
|
||||
} else if bb >= aa && bb >= ab {
|
||||
3
|
||||
} else {
|
||||
0
|
||||
};
|
||||
let mut mx = if aa > ab { aa } else { ab };
|
||||
let mut scnd = if aa > ab { ab } else { aa };
|
||||
let b_gt_mx = bb > mx;
|
||||
scnd = if b_gt_mx { mx } else if bb > scnd { bb } else { scnd };
|
||||
mx = if b_gt_mx { bb } else { mx };
|
||||
let ap_lod = math_log(mx / (tiny_d + scnd), true);
|
||||
let score_ap = smf(ap_lod as f32, 0.0, 2.0);
|
||||
|
||||
// sub-scores
|
||||
let score_r1 = smf(r, 0.0, 0.1);
|
||||
let nclus = nc.iter().filter(|&&v| v > 0).count();
|
||||
let misclust: f32 = if nclus == 1 && nc[1] == 0 {
|
||||
0.7
|
||||
} else if nclus != 3 {
|
||||
0.95
|
||||
} else {
|
||||
1.0
|
||||
};
|
||||
let g = (iap as i64 - 1).clamp(0, 2) as usize;
|
||||
let (gm_t, gd_t, gm_r, gd_r) = (tm[g], ftd[g], rm[g], frd[g]);
|
||||
let st_aa = zmf(t, gm_t + 2.0 * gd_t, gm_t + 6.0 * gd_t);
|
||||
let st_ab = zmf(((t - gm_t) / gd_t).abs(), 2.0, 6.0);
|
||||
let st_bb = smf(t, gm_t - 6.0 * gd_t, gm_t - 2.0 * gd_t);
|
||||
let mut score_t = match iap {
|
||||
1 => st_aa,
|
||||
2 => st_ab,
|
||||
3 => st_bb,
|
||||
_ => 1.0,
|
||||
};
|
||||
let r2 = smf(r, 0.0, gm_r / 10.0);
|
||||
let r3 = smf(r, gm_r - 6.0 * gd_r, gm_r - 2.0 * gd_r);
|
||||
let r4 = zmf(r, gm_r + 6.0 * gd_r, gm_r + 12.0 * gd_r);
|
||||
let mut score_r = if iap > 0 { ((score_r1 * r2) * r3) * r4 } else { 0.0 };
|
||||
let mut score_n: f32 = 1.0;
|
||||
if gt[i].is_nan() {
|
||||
score_t = f32::NAN;
|
||||
score_r = f32::NAN;
|
||||
}
|
||||
if !ok_t {
|
||||
score_t = f32::NAN;
|
||||
score_r = f32::NAN;
|
||||
score_n = f32::NAN;
|
||||
}
|
||||
let mut prelim = ((((score_ap * gt[i]) * score_t) * score_r) * score_n) * misclust;
|
||||
if r < ithr[i] {
|
||||
prelim = f32::NAN;
|
||||
iap = 0;
|
||||
}
|
||||
|
||||
// gencall_score_map
|
||||
let x = prelim as f64;
|
||||
let q = (x - 1.08) / 0.504;
|
||||
let gbell = 1.0 / (1.0 + math_pow(q * q, 1.71));
|
||||
let mut score = (math_pow(x, 0.35) * gbell) as f32;
|
||||
if score.is_nan() {
|
||||
score = 0.0;
|
||||
}
|
||||
if rz[i] || !pres[i] {
|
||||
score = 0.0;
|
||||
iap = 0;
|
||||
}
|
||||
geno[i] = if score < cutoff || io[i] { 0 } else { iap };
|
||||
scores[i] = score;
|
||||
}
|
||||
Ok((geno.into_pyarray(py), scores.into_pyarray(py)))
|
||||
}
|
||||
|
||||
/// BAF and LRR (float32, canonical NaN 0xFFC00000 where undefined).
|
||||
#[pyfunction]
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
pub fn baf_lrr<'py>(
|
||||
py: Python<'py>,
|
||||
theta: PyReadonlyArray1<'py, f32>,
|
||||
r: PyReadonlyArray1<'py, f32>,
|
||||
raw_zero: PyReadonlyArray1<'py, bool>,
|
||||
present: PyReadonlyArray1<'py, bool>,
|
||||
theta_mean: PyReadonlyArray1<'py, f32>,
|
||||
r_mean: PyReadonlyArray1<'py, f32>,
|
||||
mean_r: PyReadonlyArray1<'py, f32>,
|
||||
intensity_only: PyReadonlyArray1<'py, bool>,
|
||||
) -> PyResult<(Bound<'py, PyArray1<f32>>, Bound<'py, PyArray1<f32>>)> {
|
||||
let (th, rr, rz, pres) = (theta.as_slice()?, r.as_slice()?, raw_zero.as_slice()?, present.as_slice()?);
|
||||
let (tmv, rmv, mr, io) = (theta_mean.as_slice()?, r_mean.as_slice()?, mean_r.as_slice()?, intensity_only.as_slice()?);
|
||||
let n = th.len();
|
||||
for (l, w) in [(rr.len(), "r"), (rz.len(), "raw_zero"), (pres.len(), "present"), (mr.len(), "mean_r"), (io.len(), "intensity_only")] {
|
||||
check(l, n, w)?;
|
||||
}
|
||||
check(tmv.len(), 3 * n, "theta_mean")?;
|
||||
check(rmv.len(), 3 * n, "r_mean")?;
|
||||
let mut baf = vec![0f32; n];
|
||||
let mut lrr = vec![0f32; n];
|
||||
for i in 0..n {
|
||||
let (t, r) = (th[i], rr[i]);
|
||||
let (at, bt, ct) = (tmv[3 * i], tmv[3 * i + 1], tmv[3 * i + 2]);
|
||||
let (ar, br, cr) = (rmv[3 * i], rmv[3 * i + 1], rmv[3 * i + 2]);
|
||||
let (eq, lo, hi) = (t == bt, t < bt, t > bt);
|
||||
let (r_ref, mut b) = if eq {
|
||||
(br, 0.5f32)
|
||||
} else if lo {
|
||||
(ar + ((t - at) * (ar - br)) / (at - bt), ((t - at) / (bt - at)) * 0.5)
|
||||
} else if hi {
|
||||
(br + ((t - bt) * (cr - br)) / (ct - bt), 0.5 + ((t - bt) / (ct - bt)) * 0.5)
|
||||
} else {
|
||||
(f32::NAN, NEG_NAN32)
|
||||
};
|
||||
let denom = if io[i] { mr[i] } else { r_ref };
|
||||
let ratio = (r / denom) as f64;
|
||||
let mut l = math_log(ratio, false) as f32;
|
||||
if r == 0.0 {
|
||||
l = -f32::MAX;
|
||||
}
|
||||
if !(eq || lo || hi) {
|
||||
l = NEG_NAN32;
|
||||
}
|
||||
b = if b < 0.0 {
|
||||
0.0
|
||||
} else if b > 1.0 {
|
||||
1.0
|
||||
} else {
|
||||
b
|
||||
};
|
||||
if rz[i] || !pres[i] || b.is_nan() {
|
||||
b = NEG_NAN32;
|
||||
}
|
||||
if rz[i] || !pres[i] || l.is_nan() {
|
||||
l = NEG_NAN32;
|
||||
}
|
||||
baf[i] = b;
|
||||
lrr[i] = l;
|
||||
}
|
||||
Ok((baf.into_pyarray(py), lrr.into_pyarray(py)))
|
||||
}
|
||||
@@ -0,0 +1,226 @@
|
||||
//! Compiled kernels for SimurgArray (plan §18 P9, §19).
|
||||
//!
|
||||
//! Every kernel is a drop-in replacement for a pure-Python/NumPy reference in the `simurg` package and returns
|
||||
//! bit-identical results (checked by `tests/unit/test_core_kernels.py`):
|
||||
//!
|
||||
//! * `exp`, `log10`, `log2`, `pow` — element-wise glibc libm calls with the error semantics of CPython's `math`
|
||||
//! module as used by `simurg.genotype.gencall._c_call` (NaN input returned as is, a domain error gives the
|
||||
//! canonical +NaN, `log10(0)` = −inf, overflow = +inf). The libm symbols are the ones CPython's `math` calls, so
|
||||
//! results equal the C reference implementations (gtc2vcf, MIT) that the Python port follows.
|
||||
//! * `closest_binned` — the per-bin candidate scan of the vendor's binned nearest-neighbour search
|
||||
//! (`simurg.signal.normalization.fit._closest_binned`): float32 squared distances, first minimum wins.
|
||||
//!
|
||||
//! * `call_genotypes`, `baf_lrr` — fused per-locus GenCall calling/scoring and BAF/LRR (`gencall.rs`).
|
||||
//! * `bpm_locus_entries` — the BPM LocusEntry parser (`simurg.formats.bpm._parse_locus_entries`).
|
||||
//!
|
||||
//! No floating-point contraction (FMA) happens: Rust never fuses `a*b+c` implicitly.
|
||||
|
||||
mod bpm;
|
||||
mod gencall;
|
||||
|
||||
use numpy::{IntoPyArray, PyArray1, PyReadonlyArray1};
|
||||
use pyo3::exceptions::PyValueError;
|
||||
use pyo3::prelude::*;
|
||||
|
||||
#[link(name = "m")]
|
||||
unsafe extern "C" {
|
||||
#[link_name = "exp"]
|
||||
fn c_exp(x: f64) -> f64;
|
||||
#[link_name = "log10"]
|
||||
fn c_log10(x: f64) -> f64;
|
||||
#[link_name = "log2"]
|
||||
fn c_log2(x: f64) -> f64;
|
||||
#[link_name = "pow"]
|
||||
fn c_pow(x: f64, y: f64) -> f64;
|
||||
}
|
||||
|
||||
/// CPython's `float('nan')`.
|
||||
const PY_NAN: f64 = f64::from_bits(0x7ff8_0000_0000_0000);
|
||||
|
||||
#[inline]
|
||||
pub(crate) fn math_exp(v: f64) -> f64 {
|
||||
if v.is_nan() {
|
||||
return v;
|
||||
}
|
||||
let r = unsafe { c_exp(v) };
|
||||
if r.is_nan() { PY_NAN } else if r.is_infinite() && v.is_finite() { f64::INFINITY } else { r }
|
||||
}
|
||||
|
||||
/// CPython `m_log10`/`m_log2` + `_c_call`: x > 0 → libm; 0 → −inf (log10) or NaN (log2); x < 0, −inf → NaN.
|
||||
#[inline]
|
||||
pub(crate) fn math_log(v: f64, ten: bool) -> f64 {
|
||||
if v.is_nan() {
|
||||
return v;
|
||||
}
|
||||
if v.is_infinite() {
|
||||
return if v > 0.0 { v } else { PY_NAN };
|
||||
}
|
||||
if v > 0.0 {
|
||||
return unsafe { if ten { c_log10(v) } else { c_log2(v) } };
|
||||
}
|
||||
if v == 0.0 && ten { f64::NEG_INFINITY } else { PY_NAN }
|
||||
}
|
||||
|
||||
/// CPython `math.pow(v, e)` for a finite exponent e > 0 that is not an integer, + `_c_call`.
|
||||
#[inline]
|
||||
pub(crate) fn math_pow(v: f64, e: f64) -> f64 {
|
||||
if v.is_nan() {
|
||||
return v;
|
||||
}
|
||||
if v.is_infinite() {
|
||||
return f64::INFINITY;
|
||||
}
|
||||
if v == 0.0 {
|
||||
return 0.0;
|
||||
}
|
||||
let r = unsafe { c_pow(v, e) };
|
||||
if r.is_nan() { PY_NAN } else if r.is_infinite() { f64::INFINITY } else { r }
|
||||
}
|
||||
|
||||
fn map_f64<'py>(py: Python<'py>, x: PyReadonlyArray1<'py, f64>, f: impl Fn(f64) -> f64) -> PyResult<Bound<'py, PyArray1<f64>>> {
|
||||
let s = x.as_slice()?;
|
||||
let out: Vec<f64> = s.iter().map(|&v| f(v)).collect();
|
||||
Ok(out.into_pyarray(py))
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
fn exp<'py>(py: Python<'py>, x: PyReadonlyArray1<'py, f64>) -> PyResult<Bound<'py, PyArray1<f64>>> {
|
||||
map_f64(py, x, math_exp)
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
fn log10<'py>(py: Python<'py>, x: PyReadonlyArray1<'py, f64>) -> PyResult<Bound<'py, PyArray1<f64>>> {
|
||||
map_f64(py, x, |v| math_log(v, true))
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
fn log2<'py>(py: Python<'py>, x: PyReadonlyArray1<'py, f64>) -> PyResult<Bound<'py, PyArray1<f64>>> {
|
||||
map_f64(py, x, |v| math_log(v, false))
|
||||
}
|
||||
|
||||
#[pyfunction]
|
||||
fn pow<'py>(py: Python<'py>, x: PyReadonlyArray1<'py, f64>, e: f64) -> PyResult<Bound<'py, PyArray1<f64>>> {
|
||||
if !(e.is_finite() && e > 0.0 && e.fract() != 0.0) {
|
||||
return Err(PyValueError::new_err("pow kernel: exponent must be finite, positive and not an integer"));
|
||||
}
|
||||
map_f64(py, x, |v| math_pow(v, e))
|
||||
}
|
||||
|
||||
/// For each sweep point `p` searching bin `choice[p]` (0..11 regular, 12..22 shifted; any other value: none), the
|
||||
/// index of the closest candidate among `eligible[i]` with `b1[i] == choice` (regular) or `b2[i] == choice − 12`
|
||||
/// (shifted), in ascending order; squared distance in float32; −1 if no candidate or the best is not < 1e20.
|
||||
#[pyfunction]
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
fn closest_binned<'py>(
|
||||
py: Python<'py>,
|
||||
raw_x: PyReadonlyArray1<'py, f32>,
|
||||
raw_y: PyReadonlyArray1<'py, f32>,
|
||||
axis_x: PyReadonlyArray1<'py, f32>,
|
||||
axis_y: PyReadonlyArray1<'py, f32>,
|
||||
choice: PyReadonlyArray1<'py, i64>,
|
||||
eligible: PyReadonlyArray1<'py, i64>,
|
||||
b1: PyReadonlyArray1<'py, i64>,
|
||||
b2: PyReadonlyArray1<'py, i64>,
|
||||
) -> PyResult<Bound<'py, PyArray1<i64>>> {
|
||||
let (rx, ry) = (raw_x.as_slice()?, raw_y.as_slice()?);
|
||||
let (ax, ay, ch) = (axis_x.as_slice()?, axis_y.as_slice()?, choice.as_slice()?);
|
||||
let (el, b1, b2) = (eligible.as_slice()?, b1.as_slice()?, b2.as_slice()?);
|
||||
if ax.len() != ay.len() || ax.len() != ch.len() || el.len() != b1.len() || el.len() != b2.len() || rx.len() != ry.len() {
|
||||
return Err(PyValueError::new_err("closest_binned: length mismatch"));
|
||||
}
|
||||
let mut groups: Vec<Vec<usize>> = vec![Vec::new(); 23];
|
||||
for (i, &e) in el.iter().enumerate() {
|
||||
let e = usize::try_from(e).map_err(|_| PyValueError::new_err("closest_binned: negative index"))?;
|
||||
if e >= rx.len() {
|
||||
return Err(PyValueError::new_err("closest_binned: index out of range"));
|
||||
}
|
||||
if (0..12).contains(&b1[i]) {
|
||||
groups[b1[i] as usize].push(e);
|
||||
}
|
||||
if (0..11).contains(&b2[i]) {
|
||||
groups[12 + b2[i] as usize].push(e);
|
||||
}
|
||||
}
|
||||
let out: Vec<i64> = (0..ax.len())
|
||||
.map(|p| {
|
||||
let c = ch[p];
|
||||
if !(0..23).contains(&c) {
|
||||
return -1;
|
||||
}
|
||||
let cand = &groups[c as usize];
|
||||
let mut best_j = usize::MAX;
|
||||
let mut best = f32::INFINITY;
|
||||
for (j, &e) in cand.iter().enumerate() {
|
||||
let dx = rx[e] - ax[p];
|
||||
let dy = ry[e] - ay[p];
|
||||
let d = dx * dx + dy * dy;
|
||||
// numpy argmin: the first NaN wins, otherwise the first strict minimum
|
||||
if d.is_nan() {
|
||||
best_j = j;
|
||||
best = d;
|
||||
break;
|
||||
}
|
||||
if best_j == usize::MAX || d < best {
|
||||
best_j = j;
|
||||
best = d;
|
||||
}
|
||||
}
|
||||
if best_j == usize::MAX || !((best as f64) < 1e20) { -1 } else { cand[best_j] as i64 }
|
||||
})
|
||||
.collect();
|
||||
Ok(out.into_pyarray(py))
|
||||
}
|
||||
|
||||
/// Exact nearest neighbour (`simurg.signal.normalization.fit._closest_slow`): for each sweep point (x, y) the
|
||||
/// index of the reference point with the smallest float32 squared distance `(x-xr)²+(y-yr)²`; first minimum wins,
|
||||
/// a NaN distance wins as in `numpy.argmin`. `xref` must not be empty.
|
||||
#[pyfunction]
|
||||
fn closest_exact<'py>(
|
||||
py: Python<'py>,
|
||||
xref: PyReadonlyArray1<'py, f32>,
|
||||
yref: PyReadonlyArray1<'py, f32>,
|
||||
x: PyReadonlyArray1<'py, f32>,
|
||||
y: PyReadonlyArray1<'py, f32>,
|
||||
) -> PyResult<Bound<'py, PyArray1<i64>>> {
|
||||
let (xr, yr, xs, ys) = (xref.as_slice()?, yref.as_slice()?, x.as_slice()?, y.as_slice()?);
|
||||
if xr.is_empty() || xr.len() != yr.len() || xs.len() != ys.len() {
|
||||
return Err(PyValueError::new_err("closest_exact: empty reference or length mismatch"));
|
||||
}
|
||||
let out: Vec<i64> = xs
|
||||
.iter()
|
||||
.zip(ys)
|
||||
.map(|(&px, &py_)| {
|
||||
let mut best_j = 0usize;
|
||||
let mut best = f32::INFINITY;
|
||||
for (j, (&rx, &ry)) in xr.iter().zip(yr).enumerate() {
|
||||
let dx = px - rx;
|
||||
let dy = py_ - ry;
|
||||
let d = dx * dx + dy * dy;
|
||||
if d.is_nan() {
|
||||
return j as i64;
|
||||
}
|
||||
if j == 0 || d < best {
|
||||
best_j = j;
|
||||
best = d;
|
||||
}
|
||||
}
|
||||
best_j as i64
|
||||
})
|
||||
.collect();
|
||||
Ok(out.into_pyarray(py))
|
||||
}
|
||||
|
||||
#[pymodule]
|
||||
fn simurg_core(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
m.add("__version__", env!("CARGO_PKG_VERSION"))?;
|
||||
m.add_function(wrap_pyfunction!(exp, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(log10, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(log2, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(pow, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(closest_binned, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(closest_exact, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(bpm::bpm_locus_entries, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(gencall::call_genotypes, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(gencall::baf_lrr, m)?)?;
|
||||
Ok(())
|
||||
}
|
||||
@@ -0,0 +1,53 @@
|
||||
#!/bin/bash
|
||||
# Phase P9 performance gate on a 32-vCPU Graviton instance, GSE165845 GSA arrays (360 samples):
|
||||
# 1. build the compiled kernels (rust/simurg-core) into the venv;
|
||||
# 2. thread scaling 1 → 32 with the kernels (bench/scaling.py, per-sample phase from the run log);
|
||||
# 3. pure-Python reference at 1 and 32 threads on a subset, and byte-identity of all 360 GTCs (kernels vs pure).
|
||||
# Args: <unused data dir> <results dir>. Run with OMNI=0.
|
||||
set -uo pipefail
|
||||
OUT=$2
|
||||
W=/work/p9; R=$OUT/p9; mkdir -p $W $R
|
||||
N=$(nproc)
|
||||
|
||||
echo "=== build simurg-core $(date -u +%T) ==="
|
||||
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --profile minimal > /dev/null 2>&1
|
||||
export PATH="$HOME/.cargo/bin:$PATH"
|
||||
rustc --version
|
||||
VIRTUAL_ENV=$PWD/.venv uv pip install -q ./rust/simurg-core || { echo "!! simurg-core build failed"; exit 1; }
|
||||
.venv/bin/python -c "from simurg._core import backend; print('backend', backend())"
|
||||
.venv/bin/python -m pytest tests/unit/test_core_kernels.py -q -p no:cacheprovider 2>&1 | tail -2
|
||||
|
||||
echo "=== data $(date -u +%T) ==="
|
||||
GSA=https://webdata.illumina.com/downloads/productfiles/global-screening-array/v1-0
|
||||
mkdir -p $W/gsa && cd $W/gsa
|
||||
for z in infinium-global-screening-array-24-v1-0-c2-manifest-file-bpm-build38.zip \
|
||||
infinium-global-screening-array-v1-0-c1-cluster-file.zip; do
|
||||
curl -sSfL --retry 5 -O "$GSA/$z" && unzip -oq $z
|
||||
done
|
||||
cd - > /dev/null
|
||||
BPM=$W/gsa/GSA-24v1-0_C2.bpm; EGT=$W/gsa/GSA-24v1-0_C1_ClusterFile.egt
|
||||
.venv/bin/python scripts/p8/prepare.py --tsv scripts/p8/gse165845_samples.tsv --dl $W/geo --out $W/idats --workers 24 \
|
||||
|| exit 1
|
||||
|
||||
echo "=== scaling with kernels $(date -u +%T) ==="
|
||||
.venv/bin/python bench/scaling.py --bpm $BPM --egt $EGT --idats $W/idats --out $W/rust \
|
||||
--threads 1,2,4,8,16,$N 2>&1 | grep -v WARNING
|
||||
cp $W/rust/scaling.json $R/scaling_rust.json
|
||||
|
||||
echo "=== pure Python reference $(date -u +%T) ==="
|
||||
.venv/bin/python bench/scaling.py --bpm $BPM --egt $EGT --idats $W/idats --out $W/python --threads 1 --samples 64 \
|
||||
--pure-python 2>&1 | grep -v WARNING
|
||||
cp $W/python/scaling.json $R/scaling_python_t1.json
|
||||
.venv/bin/python bench/scaling.py --bpm $BPM --egt $EGT --idats $W/idats --out $W/python_all --threads $N \
|
||||
--pure-python 2>&1 | grep -v WARNING
|
||||
cp $W/python_all/scaling.json $R/scaling_python_t$N.json
|
||||
|
||||
echo "=== identity: 360 GTCs, kernels vs pure Python $(date -u +%T) ==="
|
||||
n=0; d=0
|
||||
for f in $(cd $W/rust/t$N && find samples -name '*.gtc'); do
|
||||
n=$((n+1)); cmp -s $W/rust/t$N/$f $W/python_all/t$N/$f || { d=$((d+1)); echo "differs: $f"; }
|
||||
done
|
||||
s=$(cmp -s $W/rust/t$N/sample_summary.tsv $W/python_all/t$N/sample_summary.tsv && echo identical || echo DIFFERS)
|
||||
echo "{\"gtcs\": $n, \"differ\": $d, \"sample_summary\": \"$s\"}" | tee $R/identity.json
|
||||
lscpu | grep -i "model name\|^CPU(s)" > $R/cpu.txt; free -g >> $R/cpu.txt
|
||||
echo "=== task done $(date -u +%T) ==="
|
||||
@@ -0,0 +1,24 @@
|
||||
"""Optional compiled kernels (``rust/simurg-core``, plan §18 P9).
|
||||
|
||||
``core`` is the ``simurg_core`` extension module when it is installed, else ``None``; every caller keeps a pure
|
||||
Python/NumPy path with bit-identical results. ``SIMURG_PURE_PYTHON=1`` forces the pure paths (tests, debugging).
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
from types import ModuleType
|
||||
|
||||
core: ModuleType | None
|
||||
if os.environ.get("SIMURG_PURE_PYTHON", "") not in ("", "0"):
|
||||
core = None
|
||||
else:
|
||||
try:
|
||||
import simurg_core as core # type: ignore[no-redef]
|
||||
except ImportError: # pragma: no cover - depends on the installation
|
||||
core = None
|
||||
|
||||
|
||||
def backend() -> str:
|
||||
"""``"rust <version>"`` or ``"python"`` (recorded in run manifests)."""
|
||||
return f"rust {core.__version__}" if core is not None else "python"
|
||||
@@ -32,6 +32,7 @@ from typing import Any
|
||||
import numpy as np
|
||||
import numpy.typing as npt
|
||||
|
||||
from simurg._core import core
|
||||
from simurg.core.errors import FormatError
|
||||
from simurg.formats.binutil import BinaryReader, load_bytes
|
||||
|
||||
@@ -223,7 +224,9 @@ def read_bpm(path: str | Path) -> Manifest:
|
||||
|
||||
|
||||
def _parse_locus_entries(r: BinaryReader, n: int, name_to_idx: dict[str, int]) -> dict[str, Any]:
|
||||
"""Hot loop: parse ``n`` LocusEntry records (optimised pure Python; Rust in phase P9)."""
|
||||
"""Hot loop: parse ``n`` LocusEntry records (compiled in ``simurg_core`` when installed; same results)."""
|
||||
if core is not None:
|
||||
return _parse_locus_entries_core(r, n, name_to_idx)
|
||||
view = memoryview(r._buf)
|
||||
size = r.size
|
||||
pos = r.pos
|
||||
@@ -416,6 +419,53 @@ def _parse_locus_entries(r: BinaryReader, n: int, name_to_idx: dict[str, int]) -
|
||||
}
|
||||
|
||||
|
||||
_STR_COLS = (
|
||||
"ilmn_ids",
|
||||
"snps",
|
||||
"chroms",
|
||||
"ilmn_strands",
|
||||
"source_strands",
|
||||
"ref_strands",
|
||||
"ploidies",
|
||||
"species",
|
||||
"genome_builds",
|
||||
"sources",
|
||||
"source_versions",
|
||||
"customer_strands",
|
||||
"source_seqs",
|
||||
"top_genomic_seqs",
|
||||
"allele_a_probe_seqs",
|
||||
"allele_b_probe_seqs",
|
||||
)
|
||||
|
||||
|
||||
def _parse_locus_entries_core(r: BinaryReader, n: int, name_to_idx: dict[str, int]) -> dict[str, Any]:
|
||||
"""``_parse_locus_entries`` through ``simurg_core.bpm_locus_entries`` (identical columns and error messages)."""
|
||||
assert core is not None
|
||||
pos = r.pos
|
||||
try:
|
||||
lists, map_infos, a, b, assay, exp_cl, int_only, versions, fracs, used = core.bpm_locus_entries(
|
||||
bytes(r._view[pos:]), n, list(name_to_idx)
|
||||
)
|
||||
except ValueError as e:
|
||||
raise r.fail(str(e)) from None
|
||||
out: dict[str, Any] = dict(zip(_STR_COLS, lists, strict=True))
|
||||
for i in np.flatnonzero(assay < 0):
|
||||
assay[i] = infer_assay_type(out["allele_a_probe_seqs"][i], out["allele_b_probe_seqs"][i], out["source_seqs"][i])
|
||||
r.pos = pos + used
|
||||
return {
|
||||
**out,
|
||||
"map_infos": map_infos,
|
||||
"address_a": a,
|
||||
"address_b": b,
|
||||
"assay_types": assay.astype(np.uint8),
|
||||
"exp_clusters": exp_cl,
|
||||
"intensity_only": int_only,
|
||||
"entry_versions": versions,
|
||||
"base_fractions": fracs,
|
||||
}
|
||||
|
||||
|
||||
_IUPAC = {
|
||||
"A": 1,
|
||||
"C": 2,
|
||||
|
||||
@@ -11,6 +11,7 @@ import math
|
||||
import numpy as np
|
||||
import numpy.typing as npt
|
||||
|
||||
from simurg._core import core
|
||||
from simurg.genotype.clusters import ClusterArrays
|
||||
from simurg.genotype.gencall import _libm
|
||||
|
||||
@@ -22,6 +23,17 @@ NEG_NAN = np.frombuffer(np.uint32(0xFFC00000).tobytes(), dtype=np.float32)[0] #
|
||||
def baf_lrr(
|
||||
theta: F32Arr, r: F32Arr, raw_zero: npt.NDArray[np.bool_], c: ClusterArrays, intensity_only: npt.NDArray[np.bool_]
|
||||
) -> tuple[F32Arr, F32Arr]:
|
||||
if core is not None:
|
||||
return core.baf_lrr( # type: ignore[no-any-return]
|
||||
np.ascontiguousarray(theta, np.float32),
|
||||
np.ascontiguousarray(r, np.float32),
|
||||
np.ascontiguousarray(raw_zero, np.bool_),
|
||||
np.ascontiguousarray(c.present, np.bool_),
|
||||
np.ascontiguousarray(c.theta_mean, np.float32).ravel(),
|
||||
np.ascontiguousarray(c.r_mean, np.float32).ravel(),
|
||||
np.ascontiguousarray(c.mean_r, np.float32),
|
||||
np.ascontiguousarray(intensity_only, np.bool_),
|
||||
)
|
||||
t = np.asarray(theta, np.float32)
|
||||
rr = np.asarray(r, np.float32)
|
||||
at, bt, ct = c.theta_mean[:, 0], c.theta_mean[:, 1], c.theta_mean[:, 2]
|
||||
|
||||
@@ -15,6 +15,7 @@ from dataclasses import dataclass
|
||||
import numpy as np
|
||||
import numpy.typing as npt
|
||||
|
||||
from simurg._core import core
|
||||
from simurg.genotype.clusters import ClusterArrays
|
||||
|
||||
F32 = np.float32
|
||||
@@ -42,8 +43,17 @@ def _c_call(fn, v: float) -> float: # type: ignore[no-untyped-def]
|
||||
return math.inf
|
||||
|
||||
|
||||
_CORE_FN = {math.exp: "exp", math.log10: "log10", math.log2: "log2"}
|
||||
|
||||
|
||||
def _libm(fn, x: F64Arr) -> F64Arr: # type: ignore[no-untyped-def]
|
||||
"""Element-wise libm call (bit-identical to C)."""
|
||||
"""Element-wise libm call (bit-identical to C); compiled when ``simurg_core`` is installed."""
|
||||
if core is not None:
|
||||
xa = np.ascontiguousarray(x, dtype=np.float64).ravel()
|
||||
if (e := getattr(fn, "exponent", None)) is not None:
|
||||
return core.pow(xa, e).reshape(np.shape(x)) # type: ignore[no-any-return]
|
||||
if (name := _CORE_FN.get(fn)) is not None:
|
||||
return getattr(core, name)(xa).reshape(np.shape(x)) # type: ignore[no-any-return]
|
||||
return np.fromiter((_c_call(fn, v) for v in x.tolist()), dtype=np.float64, count=len(x))
|
||||
|
||||
|
||||
@@ -51,6 +61,7 @@ def _pow(e: float): # type: ignore[no-untyped-def]
|
||||
def f(v: float) -> float:
|
||||
return math.pow(v, e)
|
||||
|
||||
f.exponent = e # type: ignore[attr-defined]
|
||||
return f
|
||||
|
||||
|
||||
@@ -137,6 +148,23 @@ def call_genotypes(
|
||||
cutoff: float = DEFAULT_CUTOFF,
|
||||
) -> Calls:
|
||||
"""Bayesian calls and GenCall scores for one sample (plan §7.5)."""
|
||||
if core is not None:
|
||||
f32 = [np.ascontiguousarray(a, np.float32).ravel() for a in (c.theta_mean, c.theta_dev, c.r_mean, c.r_dev)]
|
||||
geno, score = core.call_genotypes(
|
||||
np.ascontiguousarray(theta, np.float32),
|
||||
np.ascontiguousarray(r, np.float32),
|
||||
np.ascontiguousarray(raw_zero, np.bool_),
|
||||
np.ascontiguousarray(c.present, np.bool_),
|
||||
np.ascontiguousarray(c.n, np.int32).ravel(),
|
||||
*f32,
|
||||
np.ascontiguousarray(c.intensity_threshold, np.float32),
|
||||
np.ascontiguousarray(c.gentrain, np.float32),
|
||||
np.ascontiguousarray(intensity_only, np.bool_),
|
||||
F32(cutoff),
|
||||
)
|
||||
return Calls(
|
||||
genotypes=geno, scores=score, num_calls=int((geno > 0).sum()), num_intensity_only=int(intensity_only.sum())
|
||||
)
|
||||
n = len(theta)
|
||||
t = np.asarray(theta, np.float32)
|
||||
rr = np.asarray(r, np.float32)
|
||||
|
||||
@@ -160,7 +160,7 @@ def _process(job: Job) -> dict[str, Any]:
|
||||
try:
|
||||
grn, red = read_idat(job.grn), read_idat(job.red)
|
||||
check_pair(grn, red)
|
||||
_, notes = check_coverage(grn, ctx.manifest)
|
||||
_, notes = check_coverage(grn, ctx.manifest, ctx.cache)
|
||||
row["notes"].extend(notes)
|
||||
g = genotype_sample(grn, red, ctx, params, _W["started"])
|
||||
write_gtc(g, Path(job.gtc_path))
|
||||
|
||||
@@ -14,6 +14,7 @@ from dataclasses import dataclass
|
||||
import numpy as np
|
||||
import numpy.typing as npt
|
||||
|
||||
from simurg._core import core
|
||||
from simurg.signal.normalization.numerics import (
|
||||
F32,
|
||||
TINY,
|
||||
@@ -86,6 +87,8 @@ def closest_points(xref: FloatArr, yref: FloatArr, x: FloatArr, y: FloatArr) ->
|
||||
|
||||
def _closest_slow(xref: FloatArr, yref: FloatArr, x: FloatArr, y: FloatArr) -> IdxArr:
|
||||
"""Exact search: squared distances in float32; the first minimum wins (C strict ``<``)."""
|
||||
if core is not None and len(xref) and all(a.dtype == np.float32 for a in (xref, yref, x, y)):
|
||||
return core.closest_exact(*(np.ascontiguousarray(a) for a in (xref, yref, x, y))) # type: ignore[no-any-return]
|
||||
dx = x[:, None] - xref[None, :]
|
||||
dy = y[:, None] - yref[None, :]
|
||||
return np.argmin(dx * dx + dy * dy, axis=1).astype(np.int64)
|
||||
@@ -118,6 +121,12 @@ def _closest_binned(raw_x: FloatArr, raw_y: FloatArr, axis_x: FloatArr, axis_y:
|
||||
low = rem < F32(0.25)
|
||||
choice = np.where(mid, k, np.where(low, np.where(k == 0, k, 12 + k - 1), np.where(k == 11, k, 12 + k)))
|
||||
|
||||
if core is not None and all(a.dtype == np.float32 for a in (raw_x, raw_y, axis_x, axis_y)):
|
||||
i64 = np.int64
|
||||
return core.closest_binned( # type: ignore[no-any-return]
|
||||
*(np.ascontiguousarray(a, np.float32) for a in (raw_x, raw_y, axis_x, axis_y)),
|
||||
*(np.ascontiguousarray(a, i64) for a in (choice, eligible, b1, b2)),
|
||||
)
|
||||
out = np.full(len(axis_a), -1, dtype=np.int64)
|
||||
for c in np.unique(choice):
|
||||
cand = eligible[b1 == c] if c < 12 else eligible[b2 == c - 12]
|
||||
|
||||
@@ -90,9 +90,11 @@ def address_coverage(idat: Idat, manifest: Manifest) -> float:
|
||||
return float(found.mean())
|
||||
|
||||
|
||||
def check_coverage(idat: Idat, manifest: Manifest) -> tuple[float, list[tuple[str, str]]]:
|
||||
def check_coverage(
|
||||
idat: Idat, manifest: Manifest, cache: LocusMapCache | None = None
|
||||
) -> tuple[float, list[tuple[str, str]]]:
|
||||
"""IN-027: < 0.99 → E_IDAT_MANIFEST_MISMATCH (raised); < 1 → W_IDAT_LOW_ADDRESS_COVERAGE (returned)."""
|
||||
cov = address_coverage(idat, manifest)
|
||||
cov = cache.coverage(idat) if cache is not None and cache.manifest is manifest else address_coverage(idat, manifest)
|
||||
notes: list[tuple[str, str]] = []
|
||||
if cov < COVERAGE_ERROR:
|
||||
raise SimurgError(
|
||||
@@ -149,10 +151,24 @@ class LocusMapCache:
|
||||
self.manifest = manifest
|
||||
self.caddr = control_addresses(manifest)
|
||||
self._maps: dict[bytes, LocusMap] = {}
|
||||
self._coverage: dict[bytes, float] = {}
|
||||
self.max_entries = max_entries
|
||||
|
||||
@staticmethod
|
||||
def _key(idat: Idat) -> bytes:
|
||||
return hashlib.blake2b(np.ascontiguousarray(idat.address_ids).tobytes(), digest_size=16).digest()
|
||||
|
||||
def coverage(self, idat: Idat) -> float:
|
||||
"""``address_coverage`` memoised per address layout."""
|
||||
key = self._key(idat)
|
||||
if (cov := self._coverage.get(key)) is None:
|
||||
if len(self._coverage) >= self.max_entries:
|
||||
self._coverage.pop(next(iter(self._coverage)))
|
||||
cov = self._coverage[key] = address_coverage(idat, self.manifest)
|
||||
return cov
|
||||
|
||||
def get(self, idat: Idat) -> LocusMap:
|
||||
key = hashlib.blake2b(np.ascontiguousarray(idat.address_ids).tobytes(), digest_size=16).digest()
|
||||
key = self._key(idat)
|
||||
lm = self._maps.get(key)
|
||||
if lm is None:
|
||||
if len(self._maps) >= self.max_entries:
|
||||
|
||||
@@ -0,0 +1,238 @@
|
||||
"""The compiled kernels (``rust/simurg-core``, plan §18 P9) must equal the pure-Python paths bit for bit."""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import datetime as dt
|
||||
import math
|
||||
from dataclasses import fields, replace
|
||||
|
||||
import numpy as np
|
||||
import pytest
|
||||
from tests.conftest import GSA_BPM, GSA_EGT, GSA_IDATS
|
||||
|
||||
import simurg.formats.bpm as bpm_mod
|
||||
import simurg.genotype.baf_lrr as baf_mod
|
||||
import simurg.genotype.gencall as gencall_mod
|
||||
import simurg.signal.normalization.fit as fit_mod
|
||||
from simurg._core import core
|
||||
from simurg.core.errors import FormatError
|
||||
from simurg.formats.bpm import read_bpm
|
||||
from simurg.synth.files import SynthLocus, encode_bpm
|
||||
|
||||
pytestmark = pytest.mark.skipif(core is None, reason="simurg_core extension not installed")
|
||||
|
||||
MODULES = (bpm_mod, gencall_mod, fit_mod, baf_mod)
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
def pure(monkeypatch): # type: ignore[no-untyped-def]
|
||||
"""Context switch: ``with pure(): ...`` runs the pure-Python paths."""
|
||||
|
||||
class _Pure:
|
||||
def __enter__(self) -> None:
|
||||
for m in MODULES:
|
||||
monkeypatch.setattr(m, "core", None)
|
||||
|
||||
def __exit__(self, *exc: object) -> None:
|
||||
for m in MODULES:
|
||||
monkeypatch.setattr(m, "core", core)
|
||||
|
||||
return _Pure
|
||||
|
||||
|
||||
def _bits(a: np.ndarray) -> bytes:
|
||||
return np.ascontiguousarray(a).tobytes()
|
||||
|
||||
|
||||
def _edge_values() -> np.ndarray:
|
||||
neg_nan = np.frombuffer(np.uint64(0xFFF8000000000000).tobytes(), np.float64)[0]
|
||||
payload_nan = np.frombuffer(np.uint64(0x7FF8000000001234).tobytes(), np.float64)[0]
|
||||
tiny = np.finfo(np.float64).tiny
|
||||
special = [0.0, -0.0, 1.0, -1.0, math.inf, -math.inf, math.nan, neg_nan, payload_nan, tiny, tiny / 4, -tiny,
|
||||
1e-320, 709.78, 709.79, -745.2, -746.0, 1e300, -1e300, 2.0**-1074, 0.5, 10.0, 1e-5] # fmt: skip
|
||||
rng = np.random.default_rng(7)
|
||||
rand = np.concatenate(
|
||||
[rng.normal(0, 3, 20_000), rng.uniform(-800, 800, 5_000), np.exp(rng.uniform(-700, 700, 5_000))]
|
||||
)
|
||||
return np.concatenate([np.array(special, np.float64), rand, -rand[:1000]])
|
||||
|
||||
|
||||
@pytest.mark.parametrize("fn", [math.exp, math.log10, math.log2, gencall_mod._pow(1.71), gencall_mod._pow(0.35)])
|
||||
def test_libm_kernels_bit_identical(fn, pure) -> None: # type: ignore[no-untyped-def]
|
||||
x = _edge_values()
|
||||
fast = gencall_mod._libm(fn, x)
|
||||
with pure():
|
||||
ref = gencall_mod._libm(fn, x)
|
||||
assert fast.dtype == ref.dtype == np.float64
|
||||
assert _bits(fast) == _bits(ref)
|
||||
|
||||
|
||||
def test_pow_kernel_rejects_integer_exponent() -> None:
|
||||
assert core is not None
|
||||
with pytest.raises(ValueError):
|
||||
core.pow(np.ones(3), 2.0)
|
||||
|
||||
|
||||
def test_closest_binned_identical(pure) -> None: # type: ignore[no-untyped-def]
|
||||
rng = np.random.default_rng(3)
|
||||
for trial in range(20):
|
||||
n = int(rng.integers(0, 3000))
|
||||
raw_x = rng.gamma(2.0, 400.0, n).astype(np.float32)
|
||||
raw_y = rng.gamma(1.0, 60.0 if trial % 2 else 3000.0, n).astype(np.float32)
|
||||
if trial == 5 and n:
|
||||
raw_x[: n // 3] = raw_x[0] # ties: first minimum must win
|
||||
raw_y[: n // 3] = raw_y[0]
|
||||
if trial == 7 and n:
|
||||
raw_x[n // 2] = np.nan
|
||||
m = int(rng.integers(1, 400))
|
||||
axis = np.linspace(0, float(np.float32(rng.uniform(100, 5000))), m).astype(np.float32)
|
||||
zeros = np.zeros(m, np.float32)
|
||||
ax, ay = (axis, zeros) if trial % 3 else (zeros, axis)
|
||||
fast = fit_mod._closest_binned(raw_x, raw_y, ax, ay)
|
||||
with pure():
|
||||
ref = fit_mod._closest_binned(raw_x, raw_y, ax, ay)
|
||||
assert fast.dtype == ref.dtype
|
||||
assert np.array_equal(fast, ref), trial
|
||||
|
||||
|
||||
def _loci() -> list[SynthLocus]:
|
||||
return [
|
||||
SynthLocus("rs1", "1", 100, "[A/G]", 11, 0, 0, norm_id=1, source_seq="ACGT[A/G]TTGA", probe_a="ACGTA"),
|
||||
SynthLocus("rs2", "2", 200, "[A/T]", 12, 13, 1, ilmn_strand="BOT", ref_strand="-", probe_a="AC", probe_b="AT"),
|
||||
SynthLocus("rs3", "X", 300, "[C/G]", 14, 15, 2, norm_id=99),
|
||||
SynthLocus("rsü", "Y", 400, "[T/C]", 16, 0, 0, norm_id=2, intensity_only=1),
|
||||
SynthLocus("rs5", "MT", 500, "[A/C]", 17, 0, 0, norm_id=99, source_seq="A" * 200 + "[A/C]"),
|
||||
]
|
||||
|
||||
|
||||
def _manifest_dict(m) -> dict: # type: ignore[no-untyped-def]
|
||||
out = {}
|
||||
for f in fields(m):
|
||||
v = getattr(m, f.name)
|
||||
out[f.name] = (v.dtype.str, v.shape, v.tobytes()) if isinstance(v, np.ndarray) else v
|
||||
return out
|
||||
|
||||
|
||||
@pytest.mark.parametrize("entry_version", [4, 6, 7, 8])
|
||||
@pytest.mark.parametrize("order", [None, [4, 3, 2, 1, 0]])
|
||||
def test_bpm_parser_identical(tmp_path, pure, entry_version, order) -> None: # type: ignore[no-untyped-def]
|
||||
p = tmp_path / "m.bpm"
|
||||
p.write_bytes(encode_bpm(_loci(), entry_version=entry_version, entry_order=order))
|
||||
fast = _manifest_dict(read_bpm(p))
|
||||
with pure():
|
||||
ref = _manifest_dict(read_bpm(p))
|
||||
assert fast == ref
|
||||
|
||||
|
||||
def test_bpm_parser_errors_identical(tmp_path, pure) -> None: # type: ignore[no-untyped-def]
|
||||
good = encode_bpm(_loci())
|
||||
cases = {
|
||||
"truncated": good[: len(good) - 40],
|
||||
"bad_assay": encode_bpm([SynthLocus("rs1", address_b=0, assay_type=1)]),
|
||||
"bad_map": encode_bpm([SynthLocus("rs1", pos=12345)]).replace(b"\x0512345", b"\x0512x45"),
|
||||
}
|
||||
for label, data in cases.items():
|
||||
p = tmp_path / f"{label}.bpm"
|
||||
p.write_bytes(data)
|
||||
with pytest.raises(FormatError) as fast:
|
||||
read_bpm(p)
|
||||
with pure(), pytest.raises(FormatError) as ref:
|
||||
read_bpm(p)
|
||||
assert (fast.value.code, str(fast.value)) == (ref.value.code, str(ref.value)), label
|
||||
|
||||
|
||||
@pytest.mark.data
|
||||
@pytest.mark.skipif(not GSA_BPM.exists(), reason="GSA demo data not fetched")
|
||||
def test_genotype_sample_identical(pure) -> None: # type: ignore[no-untyped-def]
|
||||
from simurg.formats.egt import read_egt
|
||||
from simurg.formats.gtc import encode_gtc
|
||||
from simurg.formats.idat import read_idat
|
||||
from simurg.genotype.clusters import align_clusters
|
||||
from simurg.genotype.sample import genotype_sample, manifest_context, params_for_preset
|
||||
from simurg.signal.raw_xy import LocusMapCache
|
||||
|
||||
def run() -> bytes:
|
||||
m = read_bpm(GSA_BPM)
|
||||
ctx = manifest_context(m, align_clusters(read_egt(GSA_EGT), m))
|
||||
ctx = replace(ctx, manifest=m.lite(), cache=LocusMapCache(m.lite()))
|
||||
stem = next(iter(sorted(GSA_IDATS.rglob("*_Grn.idat")))).as_posix()[: -len("_Grn.idat")]
|
||||
g = genotype_sample(read_idat(stem + "_Grn.idat"), read_idat(stem + "_Red.idat"), ctx,
|
||||
params_for_preset("current"), dt.datetime(2026, 1, 1)) # fmt: skip
|
||||
return encode_gtc(g)
|
||||
|
||||
fast = run()
|
||||
with pure():
|
||||
ref = run()
|
||||
assert fast == ref
|
||||
|
||||
|
||||
def test_closest_exact_identical(pure) -> None: # type: ignore[no-untyped-def]
|
||||
rng = np.random.default_rng(11)
|
||||
for trial in range(10):
|
||||
n = int(rng.integers(1, 1500))
|
||||
xr = rng.normal(500, 200, n).astype(np.float32)
|
||||
yr = rng.normal(300, 100, n).astype(np.float32)
|
||||
if trial == 2:
|
||||
xr[n // 2 :] = xr[0]
|
||||
yr[n // 2 :] = yr[0]
|
||||
if trial == 3:
|
||||
xr[n // 3] = np.nan
|
||||
m = int(rng.integers(1, 300))
|
||||
x = rng.normal(500, 200, m).astype(np.float32)
|
||||
y = rng.normal(300, 100, m).astype(np.float32)
|
||||
fast = fit_mod._closest_slow(xr, yr, x, y)
|
||||
with pure():
|
||||
ref = fit_mod._closest_slow(xr, yr, x, y)
|
||||
assert np.array_equal(fast, ref), trial
|
||||
|
||||
|
||||
def _synthetic_clusters(n: int, rng: np.random.Generator): # type: ignore[no-untyped-def]
|
||||
from simurg.genotype.clusters import ClusterArrays
|
||||
|
||||
tm = np.sort(rng.uniform(0, 1, (n, 3)), axis=1).astype(np.float32)
|
||||
td = rng.uniform(0.0, 0.08, (n, 3)).astype(np.float32)
|
||||
rm = rng.uniform(0.2, 3, (n, 3)).astype(np.float32)
|
||||
rd = rng.uniform(0.0, 0.3, (n, 3)).astype(np.float32)
|
||||
counts = rng.integers(0, 200, (n, 3)).astype(np.int32)
|
||||
counts[rng.random((n, 3)) < 0.2] = 0
|
||||
counts[: n // 50] = 0 # no samples at all
|
||||
td[rng.random((n, 3)) < 0.02] = 0.0
|
||||
tm[rng.random((n, 3)) < 0.01] = np.nan
|
||||
rm[rng.random((n, 3)) < 0.01] = np.nan
|
||||
thr = rng.uniform(0, 0.3, n).astype(np.float32)
|
||||
thr[rng.random(n) < 0.3] = np.nan
|
||||
gt = rng.uniform(0, 1, n).astype(np.float32)
|
||||
gt[rng.random(n) < 0.02] = np.nan
|
||||
return ClusterArrays(
|
||||
present=rng.random(n) > 0.01, n=counts, theta_mean=tm, theta_dev=td, r_mean=rm, r_dev=rd,
|
||||
intensity_threshold=thr, gentrain=gt, mean_r=rng.uniform(0.2, 3, n).astype(np.float32),
|
||||
) # fmt: skip
|
||||
|
||||
|
||||
def test_call_genotypes_and_baf_lrr_identical(pure) -> None: # type: ignore[no-untyped-def]
|
||||
from simurg.genotype.baf_lrr import baf_lrr
|
||||
from simurg.genotype.gencall import call_genotypes
|
||||
|
||||
rng = np.random.default_rng(5)
|
||||
n = 60_000
|
||||
c = _synthetic_clusters(n, rng)
|
||||
theta = rng.uniform(-0.05, 1.05, n).astype(np.float32)
|
||||
theta[: n // 3] = c.theta_mean[: n // 3, rng.integers(0, 3)] # exactly on a cluster centre
|
||||
r = rng.gamma(2.0, 0.6, n).astype(np.float32)
|
||||
raw_zero = rng.random(n) < 0.01
|
||||
theta[raw_zero] = np.nan
|
||||
r[raw_zero] = np.nan
|
||||
r[rng.random(n) < 0.01] = 0.0
|
||||
intensity_only = rng.random(n) < 0.02
|
||||
for cutoff in (0.15, 0.0):
|
||||
fast = call_genotypes(theta, r, raw_zero, c, intensity_only, cutoff)
|
||||
with pure():
|
||||
ref = call_genotypes(theta, r, raw_zero, c, intensity_only, cutoff)
|
||||
assert _bits(fast.genotypes) == _bits(ref.genotypes)
|
||||
assert _bits(fast.scores) == _bits(ref.scores)
|
||||
assert (fast.num_calls, fast.num_intensity_only) == (ref.num_calls, ref.num_intensity_only)
|
||||
fb, fl = baf_lrr(theta, r, raw_zero, c, intensity_only)
|
||||
with pure():
|
||||
rb, rl = baf_lrr(theta, r, raw_zero, c, intensity_only)
|
||||
assert _bits(fb) == _bits(rb)
|
||||
assert _bits(fl) == _bits(rl)
|
||||
Reference in new issue
Block a user