Builds a ranked, downloadable menu of candidate flanking SNP markers around a pathogenic variant, for PGT-M linkage and karyomapping. Syngamy then reads SNP arrays after an experiment and reports which parental genome is present, on which chromosomes, and across which regions within them. Progenitor builds a SNP array for a parent nobody genotyped, out of the haploid cells that parent produced. It works out which of the two parents it built from the Y those cells carry, writes the array out as a file usable anywhere an array goes, and calls the parental origin of every input against it.

Developed by & for the Egli Lab at Columbia University Irving Medical Center.

Research use only. Candidate markers require validation and per-family phasing in a qualified genetics laboratory. Not a clinical diagnostic.

Live at originmarker.app, and in parallel at ezrakruger.cc/originmarker.

Genotype files are read in the browser. There is no endpoint that receives one.

Running it

docker compose up -d --build

Then http://localhost:8091. No API keys are required; optional ones are listed in deploy/README-deploy.md and the app degrades gracefully without them.

Without Docker:

python -m venv .venv && .venv/bin/pip install -r requirements.txt
.venv/bin/uvicorn app.main:app --reload
cd web && npm install && npm run dev

The Syngamy analysis also runs standalone, with more than the page exposes:

python -m originmarker.origin --father sperm.txt --samples embryo1.txt embryo2.txt

Tests

export PANELBUILDER_CACHE=tests/fixtures PANELBUILDER_CACHE_TTL=0
.venv/bin/python -m pytest tests/
cd web && npm run build && for f in src/*.check.ts; do node "$f"; done

Offline, against recorded API responses in tests/fixtures/. npm run build is tsc -b && vite build, and the -b is the typecheck: bare tsc has no inputs here and exits 0 regardless.

Layout

originmarker/	the library: panel building, array reading, the HMM, the report writer. Every module carries a self-check, run as python -m originmarker.<module>
app/	FastAPI over that library. app/main.py is the only entry point
web/	the SPA. Syngamy and Progenitor run entirely in the browser; nothing is uploaded
data/	23MB of bundled deCODE recombination maps, read from disk and never re-downloaded
tests/	offline, against recorded API responses in tests/fixtures/
tools/	one-off analysis scripts, kept because their numbers are cited in the docs
audit/	the accuracy audit and its record
docs/	validation write-ups per major version
deploy/	the deployment runbook

Documentation

Method, scope, data sources and limitations: Documentation, Syngamy and Progenitor.

Version history and the bugs each release fixed: CHANGELOG.md.

Accuracy audit, 27 public arrays with bench-established answers: audit/.

License

Apache 2.0.