GenomicsCPU5 credits

Geneformer — Single-Cell Foundation Model

Zero-shot cell-type annotation from expression matrices

Geneformer is a transformer foundation model trained on 30M single cells by Chan-Zuckerberg. It classifies cell types from sparse expression matrices without any training. SciRouter is the first API aggregator hosting Geneformer alongside protein models.

$0.05

per API call

credits per call

/v1/singlecell/annotate

API endpoint

Features

Zero-shot cell-type annotation

30+ canonical cell types recognized

Marker gene signature scoring

Confidence scores per cell

Up to 100 cells × 500 genes per call

Quick Start

Geneformer — Python Examplepython

import requests

result = requests.post(
    "https://scirouter.ai/v1/singlecell/annotate",
    json={
        "expression_matrix": [[5.0, 4.5, 0.1]],
        "gene_names": ["CD4", "CD3D", "CD19"]
    },
    headers={"Authorization": "Bearer sk-sci-your-key"}
).json()
for p in result["data"]["predictions"]:
    print(f"{p['predicted_cell_type']}: {p['confidence']}")