AF2Complex predicts direct physical interactions in multimeric proteins with deep learning

Mu Gao; Davi Nakajima An; Jerry M. Parks; Jeffrey Skolnick

doi:10.1038/s41467-022-29394-2

AF2Complex predicts direct physical interactions in multimeric proteins with deep learning

Mu Gao, Davi Nakajima An, Jerry M. Parks, Jeffrey Skolnick

Molecular Biophysics

Research output: Contribution to journal › Article › peer-review

135 Scopus citations

Abstract

Accurate descriptions of protein-protein interactions are essential for understanding biological systems. Remarkably accurate atomic structures have been recently computed for individual proteins by AlphaFold2 (AF2). Here, we demonstrate that the same neural network models from AF2 developed for single protein sequences can be adapted to predict the structures of multimeric protein complexes without retraining. In contrast to common approaches, our method, AF2Complex, does not require paired multiple sequence alignments. It achieves higher accuracy than some complex protein-protein docking strategies and provides a significant improvement over AF-Multimer, a development of AlphaFold for multimeric proteins. Moreover, we introduce metrics for predicting direct protein-protein interactions between arbitrary protein pairs and validate AF2Complex on some challenging benchmark sets and the E. coli proteome. Lastly, using the cytochrome c biogenesis system I as an example, we present high-confidence models of three sought-after assemblies formed by eight members of this system.

Original language	English
Article number	1744
Journal	Nature Communications
Volume	13
Issue number	1
DOIs	https://doi.org/10.1038/s41467-022-29394-2
State	Published - Dec 2022

Funding

We thank Ada Sedova for coordinating the deployment of AlphaFold2 on Summit at Oak Ridge and critical reading of the manuscript, Ryan Prout, Subil Abraham, Wael Elwasif, N. Quentin Haas for building a Singularity container, and Mark Coletti for providing Dask scripts for running AF2. We thank Jessica Forness for proofreading the manuscript. This work was supported in part by the DOE Office of Science, Office of Biological and Environmental Research (DOE DE-SC0021303, J.S. and J.P.) and the Division of General Medical Sciences of the National Institute Health (NIH R35GM118039, J.S.). The research used resources supported in part by the Director’s Discretion Project at the Oak Ridge Leadership Computing Facility, and the Advanced Scientific Computing Research (ASCR) Leadership Computing Challenge (ALCC) program (J.S, J.P., and M.G.). We also acknowledge the computing resources provided by the Partnership for an Advanced Computing Environment (PACE) at the Georgia Institute of Technology. We thank Ada Sedova for coordinating the deployment of AlphaFold2 on Summit at Oak Ridge and critical reading of the manuscript, Ryan Prout, Subil Abraham, Wael Elwasif, N. Quentin Haas for building a Singularity container, and Mark Coletti for providing Dask scripts for running AF2. We thank Jessica Forness for proofreading the manuscript. This work was supported in part by the DOE Office of Science, Office of Biological and Environmental Research (DOE DE-SC0021303, J.S. and J.P.) and the Division of General Medical Sciences of the National Institute Health (NIH R35GM118039, J.S.). The research used resources supported in part by the Director’s Discretion Project at the Oak Ridge Leadership Computing Facility, and the Advanced Scientific Computing Research (ASCR) Leadership Computing Challenge (ALCC) program (J.S, J.P., and M.G.). We also acknowledge the computing resources provided by the Partnership for an Advanced Computing Environment (PACE) at the Georgia Institute of Technology.

Funders	Funder number
Division of General Medical Sciences
National Institutes of Health
U.S. Department of Energy	DE-SC0021303
National Institute of General Medical Sciences	R35GM118039
Office of Science
Advanced Scientific Computing Research
Biological and Environmental Research
Georgia Institute of Technology

Access to Document

10.1038/s41467-022-29394-2

Cite this

@article{c3f4fe1b2742487aa33b42ecf7847c08,

title = "AF2Complex predicts direct physical interactions in multimeric proteins with deep learning",

abstract = "Accurate descriptions of protein-protein interactions are essential for understanding biological systems. Remarkably accurate atomic structures have been recently computed for individual proteins by AlphaFold2 (AF2). Here, we demonstrate that the same neural network models from AF2 developed for single protein sequences can be adapted to predict the structures of multimeric protein complexes without retraining. In contrast to common approaches, our method, AF2Complex, does not require paired multiple sequence alignments. It achieves higher accuracy than some complex protein-protein docking strategies and provides a significant improvement over AF-Multimer, a development of AlphaFold for multimeric proteins. Moreover, we introduce metrics for predicting direct protein-protein interactions between arbitrary protein pairs and validate AF2Complex on some challenging benchmark sets and the E. coli proteome. Lastly, using the cytochrome c biogenesis system I as an example, we present high-confidence models of three sought-after assemblies formed by eight members of this system.",

author = "Mu Gao and {Nakajima An}, Davi and Parks, {Jerry M.} and Jeffrey Skolnick",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s).",

year = "2022",

month = dec,

doi = "10.1038/s41467-022-29394-2",

language = "English",

volume = "13",

journal = "Nature Communications",

issn = "2041-1723",

publisher = "Nature Research",

number = "1",

}

TY - JOUR

T1 - AF2Complex predicts direct physical interactions in multimeric proteins with deep learning

AU - Gao, Mu

AU - Nakajima An, Davi

AU - Parks, Jerry M.

AU - Skolnick, Jeffrey

PY - 2022/12

Y1 - 2022/12

N2 - Accurate descriptions of protein-protein interactions are essential for understanding biological systems. Remarkably accurate atomic structures have been recently computed for individual proteins by AlphaFold2 (AF2). Here, we demonstrate that the same neural network models from AF2 developed for single protein sequences can be adapted to predict the structures of multimeric protein complexes without retraining. In contrast to common approaches, our method, AF2Complex, does not require paired multiple sequence alignments. It achieves higher accuracy than some complex protein-protein docking strategies and provides a significant improvement over AF-Multimer, a development of AlphaFold for multimeric proteins. Moreover, we introduce metrics for predicting direct protein-protein interactions between arbitrary protein pairs and validate AF2Complex on some challenging benchmark sets and the E. coli proteome. Lastly, using the cytochrome c biogenesis system I as an example, we present high-confidence models of three sought-after assemblies formed by eight members of this system.

AB - Accurate descriptions of protein-protein interactions are essential for understanding biological systems. Remarkably accurate atomic structures have been recently computed for individual proteins by AlphaFold2 (AF2). Here, we demonstrate that the same neural network models from AF2 developed for single protein sequences can be adapted to predict the structures of multimeric protein complexes without retraining. In contrast to common approaches, our method, AF2Complex, does not require paired multiple sequence alignments. It achieves higher accuracy than some complex protein-protein docking strategies and provides a significant improvement over AF-Multimer, a development of AlphaFold for multimeric proteins. Moreover, we introduce metrics for predicting direct protein-protein interactions between arbitrary protein pairs and validate AF2Complex on some challenging benchmark sets and the E. coli proteome. Lastly, using the cytochrome c biogenesis system I as an example, we present high-confidence models of three sought-after assemblies formed by eight members of this system.

UR - http://www.scopus.com/inward/record.url?scp=85127407321&partnerID=8YFLogxK

U2 - 10.1038/s41467-022-29394-2

DO - 10.1038/s41467-022-29394-2

M3 - Article

C2 - 35365655

AN - SCOPUS:85127407321

SN - 2041-1723

VL - 13

JO - Nature Communications

JF - Nature Communications

IS - 1

M1 - 1744

ER -

AF2Complex predicts direct physical interactions in multimeric proteins with deep learning

Abstract

Funding

Access to Document

Other files and links

Fingerprint

Cite this