Promoter

Binary classification that detects promoter regions in DNA sequence. The default model (g0-promoter-2000bp) is a G0 BERT-Large encoder trained on human promoter sequences from the EPDnew database.

Input

A DNA sequence passed as sequence (inline) or a server-side handle. The default model uses a 2,000 bp context window with a 1,000 bp prediction window; a 300 bp variant is also available. Longer inputs are scanned in windows, so you can submit large regions (for example a 100 kb window around a gene).

Species-specific models cover human, Drosophila, yeast, and Arabidopsis. List them with GET /v1/tasks/promoter/models.

Output

Predicted promoter regions with per-region probability scores, returned in the {data, meta} envelope. The model supports probabilities, BED, and (on the large models) saliency and attention-map outputs.

Try it

REST API: POST /v1/tasks/promoter/predict — see the getting-started guide. MCP: ask your client to fetch a region and scan it for promoters, for example "Fetch a 100 kb window around human TP53 and scan it for promoter regions." See mcp.md.