GAME: Genomic API for Model Evaluation

GAME: Genomic API for Model Evaluation#

GAME was designed for the functional genomics community to create seamless communication across pre-trained models and genomics datasets. It is a product of the feedback from many model and dataset experts and our hope is that it allows for long-lasting benchmarking of models. Models and datasets communicate via a set of predefined protocols through REST APIs. The common protocol enables any model to communicate with any dataset (although not all combinations may make sense).

The Evaluators (dataset APIs) will make prediction requests in the standard format to the Predictors (model APIs), which then return the predictions to the Evaluator in a standard format, enabling the Evaluators to calculate the model’s performance. Each of the Evaluators and Predictors will be containerized using Apptainer.

Diagram

What’s different from existing approaches#

  • Community-driven API design, support, maintenance, and updates.

  • Containerized workflows to prevent dependency installation and compatibility problems.

  • Preserved model integrity by eliminating the need for model architectural rewrites or refactoring.

Getting Started#

Explore the API Specifications

GAME: API Specifications

Get started with pre-built containers

Get started by using pre-built GAME modules

Contribute your own modules to GAME

Building your own GAME Modules

If you use GAME for your research, please cite our preprint on bioRxiv.

Feel free to reach out via email: ishika.luthra@ubc.ca