Accelerating mechanistic discovery

Protein-protein interactions (PPIs) are integral to virtually all biological processes. In the Walter lab at Harvard Medical School, we are using the deep learning system AlphaFold-Multimer (AF-M) to systematically screen for PPIs. We focus on genome maintenance, but the approach can be used for any area of biology.

We have folded hundreds of core genome maintenance proteins with each other, yielding tens of thousands of binary complex predictions. Results are displayed as an interactive matrix that allows rapid visualization and triage of PPIs.

To aid prioritization of PPIs, we used machine learning to train a new classifier called SPOC that evaluates structure predictions. When coupled with experimental validation, structure predictions have led to new mechanistic insights in DNA replication and repair (Lim et al. and Sifri et al.).

We hope that predictomes.org will accelerate the characterization of all PPIs that underpin genome maintenance and thereby catalyze the development of new drugs against cancer and other diseases.

To receive email updates, please register here.