Making Structure Prediction of Protein Complexes Accessible to the Bioscience Community

This project focused on improving access to AI-enabled protein structure prediction methods within the bioscience community. The work supports researchers using cryo-electron microscopy and structural biology approaches through the development of new computational tools and methodologies for analysing protein-protein interactions and protein complex assemblies.

Key Outputs

• DPI-score: A deep learning-based metric for assessing protein-protein interfaces in cryo-EM derived assemblies. Preprint on bioRxiv; revised manuscript, acknowledging AIBIO-UK funding, submitted to Nature Communications.
• Stitching genomics data to protein structures: Virulence factors in non-O157 Shiga toxin-producing Escherichia coli. Preprint on bioRxiv; under review at Microbial Genomics.
• Release of the open-source DPI software tool.

Impact

The project has produced two research papers and an open-source software tool, demonstrating the application of AI methods to protein complex analysis and structural biology. By developing and sharing new approaches for assessing protein-protein interfaces and linking genomics data with protein structures, the project has provided resources that can support researchers working with cryo-EM and other structural biology datasets. The associated DPI software tool further enables wider access to these methods, helping to strengthen computational capabilities within the bioscience community.