Large protein databases reveal structural complementarity and functional locality

Paweł Szczerbiak, Lukasz M. Szydlowski, Witold Wydmański, P. Douglas Renfrew, Julia Koehler Leman, Tomasz Kosciolek

Advances in protein structure prediction have resulted in a remarkable rise in high-quality 3D models, emphasizing the need for effective computational methods to handle and explore this structural data. Our study explores structural clusters derived from the AlphaFold Protein Structure Database (AFDB), a high-quality subset of ESMAtlas, and the Microbiome Immunity Project (MIP). We generate a unified low-dimensional representation of the protein structure space, revealing that while each database occupies distinct regions, they share significant overlap in functional profiles. Key biological functions cluster within specific regions, indicating a common functional landscape across diverse datasets. To support data exploration, we developed an open-access web server. These findings provide a foundation for further research into protein sequence-structure-function relationships, including taxonomic classification, environmental influences, and functional specificity.

DOI: https://doi.org/10.1101/2024.08.14.607935

Keywords: protein structure prediction3D modelsAlphaFold Protein Structure Database (AFDB)ESMAtlasMicrobiome Immunity Project (MIP)structural clusterslow-dimensional representationfunctional profilesbiological functionsopen-access web serverprotein sequence-structure-function relationships