AlphaBridge: tools for the analysis of predicted macromolecular complexes

Artificial intelligence (AI)-powered protein structure prediction methods have revolutionised how life scientists explore macromolecular function. Using AI to predict the structure of macromolecular complexes, is gaining attention for modelling known interactions and evaluating the likelihood of proteins forming multimers or interacting with other proteins or nucleic acids. There is a growing need for tools to efficiently evaluate these predicted models. We introduce new tools that use AlphaFold3’s “predicted local-distance difference test” (pLDDT), “predicted aligned error” (PAE), and “predicted distance error” (PDE) matrices in a graph-based community clustering approach to label sequence motifs involved in binary interactions. The resulting “interaction image” is processed through a multidimensional image algorithm to cluster interacting sequence motifs in three-dimensional binary interfaces. This method allows us to present the interaction information between multiple proteins and nucleic acids back to two-dimensional space using chord diagrams. These “AlphaBridge” diagrams summarise predicted interfaces and intermolecular interactions, including prediction confidence and sequence conservation scores. They are valuable for efficient screening of predictions, ranking, and scoring the confidence of predicted interactions, prior to more detailed (and resource intensive) analysis.