The Human Predictome is a Walter Lab resource at HMS that provides predictions of protein-protein interactions across the human proteome. Built to help researchers investigate how proteins work together in biological processes, the resource uses a two-step computational pipeline based on AlphaFold-Multimer. Its screening identified approximately 16,000 high-confidence interactions, estimated at about 90% precision, and approximately 112,000 lower-confidence interactions, estimated at about 50% precision.

The site offers tools to search for proteins and explore predicted interactions, alongside resources covering topics such as genome maintenance, the H2A/H2B interactome, and EndoMAP. It also provides downloads, tutorials, analysis tools, and links to resources including STRING DB and AlphaFold 3. The Predictome presents computational predictions as a starting point for research, not as confirmed biological findings; the site cautions that all predictions require experimental validation.

Its broader aim is to support understanding of cell physiology and provide a foundation for future research into mechanisms relevant to human disease and drug development