Dataset Information

Highly accurate protein structure prediction with AlphaFold.

ABSTRACT: Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort^1-4, the structures of around 100,000 unique proteins have been determined⁵, but this represents a small fraction of the billions of known protein sequences^6,7. Structural coverage is bottlenecked by the months to years of painstaking effort required to determine a single protein structure. Accurate computational approaches are needed to address this gap and to enable large-scale structural bioinformatics. Predicting the three-dimensional structure that a protein will adopt based solely on its amino acid sequence-the structure prediction component of the 'protein folding problem'⁸-has been an important open research problem for more than 50 years⁹. Despite recent progress^10-14, existing methods fall far short of atomic accuracy, especially when no homologous structure is available. Here we provide the first computational method that can regularly predict protein structures with atomic accuracy even in cases in which no similar structure is known. We validated an entirely redesigned version of our neural network-based model, AlphaFold, in the challenging 14th Critical Assessment of protein Structure Prediction (CASP14)¹⁵, demonstrating accuracy competitive with experimental structures in a majority of cases and greatly outperforming other methods. Underpinning the latest version of AlphaFold is a novel machine learning approach that incorporates physical and biological knowledge about protein structure, leveraging multi-sequence alignments, into the design of the deep learning algorithm.

SUBMITTER: Jumper J

PROVIDER: S-EPMC8371605 | biostudies-literature | 2021 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Highly accurate protein structure prediction with AlphaFold.

Jumper John J Evans Richard R Pritzel Alexander A Green Tim T Figurnov Michael M Ronneberger Olaf O Tunyasuvunakool Kathryn K Bates Russ R Žídek Augustin A Potapenko Anna A Bridgland Alex A Meyer Clemens C Kohl Simon A A SAA Ballard Andrew J AJ Cowie Andrew A Romera-Paredes Bernardino B Nikolov Stanislav S Jain Rishub R Adler Jonas J Back Trevor T Petersen Stig S Reiman David D Clancy Ellen E Zielinski Michal M Steinegger Martin M Pacholska Michalina M Berghammer Tamas T Bodenstein Sebastian S Silver David D Vinyals Oriol O Senior Andrew W AW Kavukcuoglu Koray K Kohli Pushmeet P Hassabis Demis D

Nature 20210715 7873

Proteins are essential to life, and understanding their structure can facilitate a mechanistic understanding of their function. Through an enormous experimental effort1-4, the structures of around 100,000 unique proteins have been determined5, but this represents a small fraction of the billions of known protein sequences6,7. Structural coverage is bottlenecked by the months to years of painstaking effort required to determine a single protein structure. Accurate ...[more]

PMID: 34265844

Similar Datasets

Project description:The ability to predict a protein's three-dimensional conformation represents a crucial starting point for investigating evolutionary connections with other members of the corresponding protein family, examining interactions with other proteins, and potentially utilizing this knowledge for the purpose of rational drug design. In this work, we evaluated the feasibility of improving AlphaFold2's three-dimensional protein predictions by developing a novel pipeline (AlphaMod) that incorporates AlphaFold2 with MODELLER, a template-based modeling program. Additionally, our tool can drive a comprehensive quality assessment of the tertiary protein structure by incorporating and comparing a set of different quality assessment tools. The outcomes of selected tools are combined into a composite score (BORDASCORE) that exhibits a meaningful correlation with GDT_TS and facilitates the selection of optimal models in the absence of a reference structure. To validate AlphaMod's results, we conducted evaluations using two distinct datasets summing up to 72 targets, previously used to independently assess AlphaFold2's performance. The generated models underwent evaluation through two methods: i) averaging the GDT_TS scores across all produced structures for a single target sequence, and ii) a pairwise comparison of the best structures generated by AlphaFold2 and AlphaMod. The latter, within the unsupervised setups, shows a rising accuracy of approximately 34% over AlphaFold2. While, when considering the supervised setup, AlphaMod surpasses AlphaFold2 in 18% of the instances. Finally, there is an 11% correspondence in outcomes between the diverse methodologies. Consequently, AlphaMod's best-predicted tertiary structures in several cases exhibited a significant improvement in the accuracy of the predictions with respect to the best models obtained by AlphaFold2. This pipeline paves the way for the integration of additional data and AI-based algorithms to further improve the reliability of the predictions.

Dataset Information

Highly accurate protein structure prediction with AlphaFold.

Publications

Highly accurate protein structure prediction with AlphaFold.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets