Kemal Yaylali
e76ae847a1
A review of the ranking's arithmetic found four things wrong, all of which made the score look better informed than it was. Measurements below are from this repo, not estimates. **Components now abstain instead of inventing a number.** A run without a VEP cache returns no allele frequencies, and rarity_score(None) read that as "absent from gnomAD, therefore maximally rare" and awarded every variant a free 0.25. jobs.has_frequencies / has_effect_scores record what the run actually produced, absent components are dropped from the weighted mean, and the remaining weights are renormalised so the score keeps its meaning. The UI shows "not looked up" rather than a bar, and the funnel stops calling a step "rare" when nothing was filtered. **Allele frequency is no longer a model feature.** It dominated: the same missense variant scored 0.887 at AF 0 and 0.0003 at AF 0.01. That double- counted, because the ranking already scores frequency explicitly, putting ~45% of every rank on one measurement; and it was circular, because ACMG assigns ClinVar's benign labels using frequency (BA1/BS1). Retraining without it moves missense AUROC from 0.872 to 0.500 — exactly random. The old figure was allele frequency, not variant-effect knowledge. The model therefore abstains unless CADD or AlphaMissense is present, since otherwise it only restates the consequence class. **Phenotype matching is weighted by information content** and HPO annotations are propagated up the ontology. Counting terms alike let "global developmental delay" (IC 0.93) count as much as "dilated left subclavian artery" (IC 7.88). **A real bug in the propagation, found by checking it.** The ancestor walk read a pre-order DFS backwards, which on a DAG lets a term resolve before one of its parents and inherit that parent alone instead of its lineage. It dropped 399 terms out of the phenotype branch, Camptodactyly and Chiari malformation among them. Now a true post-order, tested against a reference transitive closure. The ontology arithmetic moved to rarelens_ml.hpo so it is covered by tests, and rarelens_ml.benchmark measures the whole thing: across 10,178 published cases the causal gene ranks first 45.9-81.0% of the time against 5,269 genes, versus 0.02% for chance. docs/data.md reports that with its contamination (HPO's annotations come from these same case reports), and includes the measurement showing information-content weighting earns its place while propagation does not - kept anyway, for a reason the docs argue rather than assume.
108 lines
3.9 KiB
Python
108 lines
3.9 KiB
Python
"""Score a job's variants with the registered MLflow model.
|
|
|
|
The registered model is a pyfunc that owns its feature engineering (rarelens_ml.features travels
|
|
with it as model code) and returns P(pathogenic). Serving therefore only sends the raw columns
|
|
below and cannot drift from training.
|
|
"""
|
|
import asyncio
|
|
import uuid
|
|
from collections.abc import Sequence
|
|
from typing import Any
|
|
|
|
import mlflow
|
|
import pandas as pd
|
|
from mlflow import MlflowClient
|
|
from sqlalchemy import func, select
|
|
from sqlalchemy.dialects.postgresql import insert
|
|
from sqlalchemy.ext.asyncio import AsyncSession
|
|
|
|
from app.config import settings
|
|
from app.models import Prediction, Variant
|
|
|
|
# Must match rarelens_ml.features.RAW_COLUMNS. Allele frequency is not among them: the ranking
|
|
# scores frequency itself, and feeding it here too counted one measurement twice.
|
|
RAW_COLUMNS = ["impact", "consequence", "cadd_phred", "am_pathogenicity"]
|
|
CHUNK_SIZE = 5000
|
|
|
|
_models: dict[str, Any] = {} # model version -> loaded pyfunc
|
|
|
|
|
|
def load_model() -> tuple[Any, str]:
|
|
"""Return (model, version), from MODEL_URI if set, else from the registry alias."""
|
|
if settings.model_uri:
|
|
return _load_uri(settings.model_uri)
|
|
client = MlflowClient(tracking_uri=settings.mlflow_tracking_uri)
|
|
version = client.get_model_version_by_alias(settings.model_name, settings.model_alias).version
|
|
if version not in _models:
|
|
mlflow.set_tracking_uri(settings.mlflow_tracking_uri)
|
|
_models.clear()
|
|
_models[version] = mlflow.pyfunc.load_model(f"models:/{settings.model_name}/{version}")
|
|
return _models[version], version
|
|
|
|
|
|
def _load_uri(uri: str) -> tuple[Any, str]:
|
|
"""Load a model artifact directly (gs://...): no tracking server, nothing running when idle."""
|
|
version = uri.rstrip("/").rsplit("/", 1)[-1][:40] or "uri"
|
|
if version not in _models:
|
|
_models.clear()
|
|
_models[version] = mlflow.pyfunc.load_model(uri)
|
|
return _models[version], version
|
|
|
|
|
|
def raw_frame(variants: Sequence[Variant]) -> pd.DataFrame:
|
|
return pd.DataFrame(
|
|
{
|
|
"impact": [v.impact for v in variants],
|
|
"consequence": [v.consequence for v in variants],
|
|
"cadd_phred": [v.annotations.get("CADD_PHRED") for v in variants],
|
|
"am_pathogenicity": [v.annotations.get("am_pathogenicity") for v in variants],
|
|
},
|
|
columns=RAW_COLUMNS,
|
|
)
|
|
|
|
|
|
async def score_job(job_id: uuid.UUID, session: AsyncSession) -> tuple[int, str]:
|
|
model, version = await asyncio.to_thread(load_model)
|
|
scored = 0
|
|
last_id = 0
|
|
# Keyset pagination keeps memory flat for whole-genome jobs.
|
|
while True:
|
|
variants = (
|
|
await session.scalars(
|
|
select(Variant)
|
|
.where(Variant.job_id == job_id, Variant.id > last_id)
|
|
.order_by(Variant.id)
|
|
.limit(CHUNK_SIZE)
|
|
)
|
|
).all()
|
|
if not variants:
|
|
break
|
|
scores = await asyncio.to_thread(model.predict, raw_frame(variants))
|
|
stmt = insert(Prediction).values(
|
|
[
|
|
{
|
|
"variant_id": v.id,
|
|
"model_name": settings.model_name,
|
|
"model_version": version,
|
|
"score": float(s),
|
|
}
|
|
for v, s in zip(variants, scores, strict=True)
|
|
]
|
|
)
|
|
# Re-scoring (e.g. after a new model version) replaces the previous prediction.
|
|
await session.execute(
|
|
stmt.on_conflict_do_update(
|
|
index_elements=[Prediction.variant_id],
|
|
set_={
|
|
"model_name": stmt.excluded.model_name,
|
|
"model_version": stmt.excluded.model_version,
|
|
"score": stmt.excluded.score,
|
|
"created_at": func.now(),
|
|
},
|
|
)
|
|
)
|
|
scored += len(variants)
|
|
last_id = variants[-1].id
|
|
await session.commit()
|
|
return scored, version
|