Mining folded proteomes in the era of accurate structure prediction

Abstract

Protein structure fundamentally underpins the function and processes of numerous biological systems. Fold recognition algorithms offer a sensitive and robust tool to detect structural, and thereby functional, similarities between distantly related homologs. In the era of accurate structure prediction owing to advances in machine learning techniques, previously curated sequence databases have become a rich source of biological information. Here, we use bioinformatic fold recognition algorithms to scan the entire AlphaFold structure database to identify novel protein family members, infer function and group predicted protein structures. As an example of the utility of this approach, we identify novel, previously unknown members of various pore-forming protein families, including MACPFs, GSDMs and aerolysin-like proteins. Further, we explore the use of structure-based mining for functional inference.

Caenorhabditis elegans MES-3 is a highly divergent ortholog of the canonical PRC2 component SUZ12

“…The here described similarities and differences between SUZ12 and MES-3 should facilitate further experiments to elucidate the specific mechanisms by which MES-3 acts in PRC2 in C. elegans . Our work joins a rapidly growing set of in silico predictions of previously undetected homologies made possible by unprecedented advances in deep-learning driven structure prediction ( Bayly-Jones and Whisstock, 2021 ; Sanchez-Pulido and Ponting, 2021 ).…”

The Statistical Trends of Protein Evolution: A Lesson from AlphaFold Database

“…The extensive structural dataset not only offers the potential to answer the critical questions about life activities and human health, but also suggests a new paradigm for studying protein evolution. Instead of focusing only on specific families of proteins (16), we can now perform comparative analyses of proteins structures in different species. The statistics of the full-proteome protein structures may indicate the trend of species evolution.…”

“…It is difficult to construct a database for analyzing the evolution of all proteins in various species since it is costly and time-consuming to determine the structures of all proteins contained within the proteome of a species by experiments.Notably, the recent development of artificial intelligence (AI) provides us with new and powerful tools to help us elucidate the trends in protein evolution on the macroscopic scale. AlphaFold, developed by DeepMind, is an artificial intelligence system based on deep learning, which performs protein structure predictions with high accuracy 14,15 and finds many applications in medical and biological sciences [16][17][18] . Although some of the predictions were noted to have limitations 19,20 , AlphaFold had already won an unprecedented and overwhelming success.…”

Caenorhabditis elegansMES-3 is a highly divergent ortholog of the canonical PRC2 component SUZ12

“…The here described similarities and differences between SUZ12 and MES-3 should facilitate further experiments to elucidate the specific mechanisms by which MES-3 acts in PRC2 in C. elegans. Our work joins a rapidly growing set of in silico predictions of previously undetected homologies made possible by unprecedented advances in deep-learning driven structure prediction 22,30 .…”

“…We capitalized on recent advantages on computational prediction approaches that enable to derive high-quality structures of protein monomers or multimers 23,24 , which enables to study protein function and evolution at unprecedented scale 22,30 . We demonstrate that MES-3 is a diverged ortholog of SUZ12, and that MES-3 may associate with MES-2, MES-6, and LIN-53, similar to the orthologous proteins in human PRC2.…”