Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance
Abstract Feature selection is essential in the analysis of molecular systems and many other fields, but several uncertainties remain: What is the optimal number of features for a simplified, interpretable model that retains essential information? How should features with different units be aligned,...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-01-01
|
| Series: | Nature Communications |
| Online Access: | https://doi.org/10.1038/s41467-024-55449-7 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841559243773181952 |
|---|---|
| author | Romina Wild Felix Wodaczek Vittorio Del Tatto Bingqing Cheng Alessandro Laio |
| author_facet | Romina Wild Felix Wodaczek Vittorio Del Tatto Bingqing Cheng Alessandro Laio |
| author_sort | Romina Wild |
| collection | DOAJ |
| description | Abstract Feature selection is essential in the analysis of molecular systems and many other fields, but several uncertainties remain: What is the optimal number of features for a simplified, interpretable model that retains essential information? How should features with different units be aligned, and how should their relative importance be weighted? Here, we introduce the Differentiable Information Imbalance (DII), an automated method to rank information content between sets of features. Using distances in a ground truth feature space, DII identifies a low-dimensional subset of features that best preserves these relationships. Each feature is scaled by a weight, which is optimized by minimizing the DII through gradient descent. This allows simultaneously performing unit alignment and relative importance scaling, while preserving interpretability. DII can also produce sparse solutions and determine the optimal size of the reduced feature space. We demonstrate the usefulness of this approach on two benchmark molecular problems: (1) identifying collective variables that describe conformations of a biomolecule, and (2) selecting features for training a machine-learning force field. These results show the potential of DII in addressing feature selection challenges and optimizing dimensionality in various applications. The method is available in the Python library DADApy. |
| format | Article |
| id | doaj-art-d2550552c05f48f1a68917fe8623f9b2 |
| institution | Kabale University |
| issn | 2041-1723 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Nature Communications |
| spelling | doaj-art-d2550552c05f48f1a68917fe8623f9b22025-01-05T12:40:24ZengNature PortfolioNature Communications2041-17232025-01-0116111210.1038/s41467-024-55449-7Automatic feature selection and weighting in molecular systems using Differentiable Information ImbalanceRomina Wild0Felix Wodaczek1Vittorio Del Tatto2Bingqing Cheng3Alessandro Laio4International School for Advanced Studies (SISSA)The Institute of Science and Technology Austria (ISTA)International School for Advanced Studies (SISSA)The Institute of Science and Technology Austria (ISTA)International School for Advanced Studies (SISSA)Abstract Feature selection is essential in the analysis of molecular systems and many other fields, but several uncertainties remain: What is the optimal number of features for a simplified, interpretable model that retains essential information? How should features with different units be aligned, and how should their relative importance be weighted? Here, we introduce the Differentiable Information Imbalance (DII), an automated method to rank information content between sets of features. Using distances in a ground truth feature space, DII identifies a low-dimensional subset of features that best preserves these relationships. Each feature is scaled by a weight, which is optimized by minimizing the DII through gradient descent. This allows simultaneously performing unit alignment and relative importance scaling, while preserving interpretability. DII can also produce sparse solutions and determine the optimal size of the reduced feature space. We demonstrate the usefulness of this approach on two benchmark molecular problems: (1) identifying collective variables that describe conformations of a biomolecule, and (2) selecting features for training a machine-learning force field. These results show the potential of DII in addressing feature selection challenges and optimizing dimensionality in various applications. The method is available in the Python library DADApy.https://doi.org/10.1038/s41467-024-55449-7 |
| spellingShingle | Romina Wild Felix Wodaczek Vittorio Del Tatto Bingqing Cheng Alessandro Laio Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance Nature Communications |
| title | Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance |
| title_full | Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance |
| title_fullStr | Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance |
| title_full_unstemmed | Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance |
| title_short | Automatic feature selection and weighting in molecular systems using Differentiable Information Imbalance |
| title_sort | automatic feature selection and weighting in molecular systems using differentiable information imbalance |
| url | https://doi.org/10.1038/s41467-024-55449-7 |
| work_keys_str_mv | AT rominawild automaticfeatureselectionandweightinginmolecularsystemsusingdifferentiableinformationimbalance AT felixwodaczek automaticfeatureselectionandweightinginmolecularsystemsusingdifferentiableinformationimbalance AT vittoriodeltatto automaticfeatureselectionandweightinginmolecularsystemsusingdifferentiableinformationimbalance AT bingqingcheng automaticfeatureselectionandweightinginmolecularsystemsusingdifferentiableinformationimbalance AT alessandrolaio automaticfeatureselectionandweightinginmolecularsystemsusingdifferentiableinformationimbalance |