Protein functions, such as enzymatic activities, binding interactions, and membrane transport, exist as islands in the “archipelago” of the protein function landscape. Machine learning (ML) algorithms have tried to bridge this gap, but are still unable to find a general solution for predicting any protein’s function from its DNA sequence.

A general solution for predicting any protein function from sequence would catalyze a transformation in the field of biology.

We propose to develop an experimental platform and unified data ontology for collecting datasets from different functional ‘islands’ to build predictive models for individual protein functions. The experimental strategy uses a pooled, growth-based assay measured with DNA sequencing to create a simple, yet adaptable system that can be easily expanded to encompass new functions.