Integrating physics in deep learning algorithms: a force field as a PyTorch module

There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.

Abstract

Motivation

Deep learning algorithms applied to structural biology often struggle to converge to meaningful solutions when limited data is available, since they are required to learn complex physical rules from examples. State-of-the-art force-fields, however, cannot interface with deep learning algorithms due to their implementation.

Results

We present MadraX, a forcefield implemented as a differentiable PyTorch module, able to interact with deep learning algorithms in an end-to-end fashion.

Availability and implementation

MadraX documentation, together with tutorials and installation guide, is available at madrax.readthedocs.io.

Related collections

Most cited references 17

Record: found
Abstract: found
Article: found

Is Open Access

Highly accurate protein structure prediction with AlphaFold

John Jumper, Richard Evans, Alexander Pritzel … (2021)

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 determined 5 , 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 years 9 . 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. AlphaFold predicts protein structures with an accuracy competitive with experimental structures in the majority of cases using a novel deep learning architecture.

0 comments Cited 11551 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: found
Article: not found

Development and testing of a general amber force field.

Junmei Wang, Romain M. Wolf, James W. Caldwell … (2004)

We describe here a general Amber force field (GAFF) for organic molecules. GAFF is designed to be compatible with existing Amber force fields for proteins and nucleic acids, and has parameters for most organic and pharmaceutical molecules that are composed of H, C, N, O, S, P, and halogens. It uses a simple functional form and a limited number of atom types, but incorporates both empirical and heuristic models to estimate force constants and partial atomic charges. The performance of GAFF in test cases is encouraging. In test I, 74 crystallographic structures were compared to GAFF minimized structures, with a root-mean-square displacement of 0.26 A, which is comparable to that of the Tripos 5.2 force field (0.25 A) and better than those of MMFF 94 and CHARMm (0.47 and 0.44 A, respectively). In test II, gas phase minimizations were performed on 22 nucleic acid base pairs, and the minimized structures and intermolecular energies were compared to MP2/6-31G* results. The RMS of displacements and relative energies were 0.25 A and 1.2 kcal/mol, respectively. These data are comparable to results from Parm99/RESP (0.16 A and 1.18 kcal/mol, respectively), which were parameterized to these base pairs. Test III looked at the relative energies of 71 conformational pairs that were used in development of the Parm99 force field. The RMS error in relative energies (compared to experiment) is about 0.5 kcal/mol. GAFF can be applied to wide range of molecules in an automatic fashion, making it suitable for rational drug design and database searching. Copyright 2004 Wiley Periodicals, Inc.

0 comments Cited 2447 times – based on 0 reviews      Review now

Bookmark

Record: found
Abstract: found
Article: not found

The FoldX web server: an online force field

Joost Schymkowitz, Jesper Ferkinghoff-Borg, François Stricher … (2005)

FoldX is an empirical force field that was developed for the rapid evaluation of the effect of mutations on the stability, folding and dynamics of proteins and nucleic acids. The core functionality of FoldX, namely the calculation of the free energy of a macromolecule based on its high-resolution 3D structure, is now publicly available through a web server at . The current release allows the calculation of the stability of a protein, calculation of the positions of the protons and the prediction of water bridges, prediction of metal binding sites and the analysis of the free energy of complex formation. Alanine scanning, the systematic truncation of side chains to alanine, is also included. In addition, some reporting functions have been added, and it is now possible to print both the atomic interaction networks that constitute the protein, print the structural and energetic details of the interactions per atom or per residue, as well as generate a general quality report of the pdb structure. This core functionality will be further extended as more FoldX applications are developed.

0 comments Cited 521 times – based on 0 reviews      Review now

Bookmark

All references

Author and article information

Contributors

Gabriele Orlando:

ORCID: https://orcid.org/0000-0002-5935-5258

Luis Serrano:

ORCID: https://orcid.org/0000-0002-5276-1392

Joost Schymkowitz

Frederic Rousseau

Lenore Cowen: Role: Associate Editor

Journal

Journal ID (nlm-ta): Bioinformatics

Journal ID (iso-abbrev): Bioinformatics

Journal ID (publisher-id): bioinformatics

Title: Bioinformatics

Publisher: Oxford University Press

ISSN (Print): 1367-4803

ISSN (Electronic): 1367-4811

Publication date Collection: April 2024

Publication date (Electronic): 21 March 2024

Publication date PMC-release: 21 March 2024

Volume: 40

Issue: 4

Electronic Location Identifier: btae160

Affiliations

Switch Laboratory, VIB Center for Brain and Disease Research, VIB , Leuven 3000, Belgium

Switch Laboratory, Department of Cellular and Molecular Medicine, KU Leuven , Leuven 3000, Belgium

Switch Laboratory, VIB Center for AI & Computational Biology, VIB , Leuven 3000, Belgium

Centre for Genomic Regulation (CRG), The Barcelona Institute for Science and Technology , Dr Aiguader 88, Barcelona 08003, Spain

Universitat Pompeu Fabra (UPF) , Barcelona, Spain

IC REA, Pg. Lluis Companys 23 , Barcelona 08010, Spain

Switch Laboratory, VIB Center for Brain and Disease Research, VIB , Leuven 3000, Belgium

Switch Laboratory, Department of Cellular and Molecular Medicine, KU Leuven , Leuven 3000, Belgium

Switch Laboratory, VIB Center for AI & Computational Biology, VIB , Leuven 3000, Belgium

Switch Laboratory, VIB Center for Brain and Disease Research, VIB , Leuven 3000, Belgium

Switch Laboratory, Department of Cellular and Molecular Medicine, KU Leuven , Leuven 3000, Belgium

Switch Laboratory, VIB Center for AI & Computational Biology, VIB , Leuven 3000, Belgium

Author notes

Corresponding authors. Switch Laboratory, Department of Cellular and Molecular Medicine, KU Leuven, Herestraat 49, Leuven, 3000, Belgium. E-mails: joost.schymkowitz@ 123456kuleuven.be (J.S.) and frederic.rousseau@ 123456kuleuven.be (F.R.)

Author information

Gabriele Orlando https://orcid.org/0000-0002-5935-5258

Luis Serrano https://orcid.org/0000-0002-5276-1392

Article

Publisher ID: btae160

DOI: 10.1093/bioinformatics/btae160

PMC ID: 11007235

PubMed ID: 38514422

SO-VID: e082b06d-525f-4b7c-b611-0b913a39c9cb

License:

This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.

History

Date received : 01 September 2023

Date revision received : 08 February 2024

Date: 13 February 2024

Date accepted : 19 March 2024

Date: 10 April 2024

Page count

Pages: 5

Funding

Funded by: Flanders Institute for Biotechnology, DOI 10.13039/501100004727;

Funded by: Fund for Scientific Research Flanders;

Award ID: S000523N

Award ID: S000722N

Funded by: Spanish Ministry of Science and Innovation through the Centro de Excelencia Severo Ochoa;

Award ID: CEX2020-001049-S

Award ID: MCIN/AEI /10.13039/501100011033

Integrating physics in deep learning algorithms: a force field as a PyTorch module

Read this article at

Abstract

Motivation

Results

Availability and implementation

Related collections

REPO4EU WP2 Databases

Most cited references 17

Highly accurate protein structure prediction with AlphaFold

Development and testing of a general amber force field.

The FoldX web server: an online force field

Author and article information

Contributors

Journal

Affiliations

Author notes

Author information

Article

History

Page count

Funding

Categories

Comments

Comment on this article

Similar content 24

Most referenced authors 1,176