Skip to main content
The Bunker Longevity catalog covers four life-science domains: protein structure, genomics, drug discovery, and imaging. Every model is served through the same API surface, so switching between them is a matter of changing the model name.

​
Install

Every model in the catalog is accessible through the bunker-fold Python package:
See Installation for full setup.

​
All models

​
Pick by task

  • Predict a protein structure: BunkerFold-3 for sub-angstrom structure prediction with full side chains.
  • Call variants from a genome: GenoScan-LLM to flag pathogenic variants from BAM, CRAM, or VCF inputs.
  • Design a small molecule: MolGen-7 to generate candidates against a specific target with ADMET constraints.
  • Reconstruct a cryo-EM map: CryoNet to resolve 3D density maps with 10x fewer particles than standard pipelines.
  • Screen for toxicity: ToxPredict-v2 to filter compound libraries before synthesis.
  • Annotate the epigenome: EpiMark to predict histone marks and methylation from sequence context alone.

​
Shared API surface

Every model uses the same three endpoints:
  • POST /v1/models/{model}/deploy starts an inference instance.
  • POST /v1/models/{model}/predict runs a prediction against a deployed instance.
  • GET /v1/models/{model}/status reports deployment health and queue depth.
See the API Reference for full request and response schemas.