Efficient interatomic descriptors for accurate machine learning force fields of extended molecules
Abstract Machine learning force fields (MLFFs) are gradually evolving towards enabling molecular dynamics simulations of molecules and materials with ab initio accuracy but at a small fraction of the computational cost. However, several challenges remain to be addressed to enable predictive MLFF sim...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2023-06-01
|
| Series: | Nature Communications |
| Online Access: | https://doi.org/10.1038/s41467-023-39214-w |
| _version_ | 1797778826812456960 |
|---|---|
| author | Adil Kabylda Valentin Vassilev-Galindo Stefan Chmiela Igor Poltavsky Alexandre Tkatchenko |
| author_facet | Adil Kabylda Valentin Vassilev-Galindo Stefan Chmiela Igor Poltavsky Alexandre Tkatchenko |
| author_sort | Adil Kabylda |
| collection | DOAJ |
| description | Abstract Machine learning force fields (MLFFs) are gradually evolving towards enabling molecular dynamics simulations of molecules and materials with ab initio accuracy but at a small fraction of the computational cost. However, several challenges remain to be addressed to enable predictive MLFF simulations of realistic molecules, including: (1) developing efficient descriptors for non-local interatomic interactions, which are essential to capture long-range molecular fluctuations, and (2) reducing the dimensionality of the descriptors to enhance the applicability and interpretability of MLFFs. Here we propose an automatized approach to substantially reduce the number of interatomic descriptor features while preserving the accuracy and increasing the efficiency of MLFFs. To simultaneously address the two stated challenges, we illustrate our approach on the example of the global GDML MLFF. We found that non-local features (atoms separated by as far as 15 Å in studied systems) are crucial to retain the overall accuracy of the MLFF for peptides, DNA base pairs, fatty acids, and supramolecular complexes. Interestingly, the number of required non-local features in the reduced descriptors becomes comparable to the number of local interatomic features (those below 5 Å). These results pave the way to constructing global molecular MLFFs whose cost increases linearly, instead of quadratically, with system size. |
| first_indexed | 2024-03-12T23:22:09Z |
| format | Article |
| id | doaj.art-4aaf3e78e81044debdb27aae42abcef1 |
| institution | Directory Open Access Journal |
| issn | 2041-1723 |
| language | English |
| last_indexed | 2024-03-12T23:22:09Z |
| publishDate | 2023-06-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Nature Communications |
| spelling | doaj.art-4aaf3e78e81044debdb27aae42abcef12023-07-16T11:21:05ZengNature PortfolioNature Communications2041-17232023-06-0114111210.1038/s41467-023-39214-wEfficient interatomic descriptors for accurate machine learning force fields of extended moleculesAdil Kabylda0Valentin Vassilev-Galindo1Stefan Chmiela2Igor Poltavsky3Alexandre Tkatchenko4Department of Physics and Materials Science, University of LuxembourgDepartment of Physics and Materials Science, University of LuxembourgMachine Learning Group, Technische Universität BerlinDepartment of Physics and Materials Science, University of LuxembourgDepartment of Physics and Materials Science, University of LuxembourgAbstract Machine learning force fields (MLFFs) are gradually evolving towards enabling molecular dynamics simulations of molecules and materials with ab initio accuracy but at a small fraction of the computational cost. However, several challenges remain to be addressed to enable predictive MLFF simulations of realistic molecules, including: (1) developing efficient descriptors for non-local interatomic interactions, which are essential to capture long-range molecular fluctuations, and (2) reducing the dimensionality of the descriptors to enhance the applicability and interpretability of MLFFs. Here we propose an automatized approach to substantially reduce the number of interatomic descriptor features while preserving the accuracy and increasing the efficiency of MLFFs. To simultaneously address the two stated challenges, we illustrate our approach on the example of the global GDML MLFF. We found that non-local features (atoms separated by as far as 15 Å in studied systems) are crucial to retain the overall accuracy of the MLFF for peptides, DNA base pairs, fatty acids, and supramolecular complexes. Interestingly, the number of required non-local features in the reduced descriptors becomes comparable to the number of local interatomic features (those below 5 Å). These results pave the way to constructing global molecular MLFFs whose cost increases linearly, instead of quadratically, with system size.https://doi.org/10.1038/s41467-023-39214-w |
| spellingShingle | Adil Kabylda Valentin Vassilev-Galindo Stefan Chmiela Igor Poltavsky Alexandre Tkatchenko Efficient interatomic descriptors for accurate machine learning force fields of extended molecules Nature Communications |
| title | Efficient interatomic descriptors for accurate machine learning force fields of extended molecules |
| title_full | Efficient interatomic descriptors for accurate machine learning force fields of extended molecules |
| title_fullStr | Efficient interatomic descriptors for accurate machine learning force fields of extended molecules |
| title_full_unstemmed | Efficient interatomic descriptors for accurate machine learning force fields of extended molecules |
| title_short | Efficient interatomic descriptors for accurate machine learning force fields of extended molecules |
| title_sort | efficient interatomic descriptors for accurate machine learning force fields of extended molecules |
| url | https://doi.org/10.1038/s41467-023-39214-w |
| work_keys_str_mv | AT adilkabylda efficientinteratomicdescriptorsforaccuratemachinelearningforcefieldsofextendedmolecules AT valentinvassilevgalindo efficientinteratomicdescriptorsforaccuratemachinelearningforcefieldsofextendedmolecules AT stefanchmiela efficientinteratomicdescriptorsforaccuratemachinelearningforcefieldsofextendedmolecules AT igorpoltavsky efficientinteratomicdescriptorsforaccuratemachinelearningforcefieldsofextendedmolecules AT alexandretkatchenko efficientinteratomicdescriptorsforaccuratemachinelearningforcefieldsofextendedmolecules |