Predicting Infrared Spectra with Message Passing Neural Networks
Infrared (IR) spectroscopy remains an important tool for chemical characterization and identification. Chemprop-IR has been developed as a software package for the prediction of IR spectra through the use of machine learning. This work serves the dual purpose of providing a trained general-purpose m...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Published: |
American Chemical Society (ACS)
2021
|
| Online Access: | https://hdl.handle.net/1721.1/131020 |
| _version_ | 1826202594370912256 |
|---|---|
| author | McGill, Charles Forsuelo, Michael Guan, Yanfei Green Jr, William H |
| author2 | Massachusetts Institute of Technology. Department of Chemical Engineering |
| author_facet | Massachusetts Institute of Technology. Department of Chemical Engineering McGill, Charles Forsuelo, Michael Guan, Yanfei Green Jr, William H |
| author_sort | McGill, Charles |
| collection | MIT |
| description | Infrared (IR) spectroscopy remains an important tool for chemical characterization and identification. Chemprop-IR has been developed as a software package for the prediction of IR spectra through the use of machine learning. This work serves the dual purpose of providing a trained general-purpose model for the prediction of IR spectra with ease and providing the Chemprop-IR software framework for the training of new models. In Chemprop-IR, molecules are encoded using a directed message passing neural network, allowing for molecule latent representations to be learned and optimized for the task of spectral predictions. Model training incorporates spectra metrics and normalization techniques that offer better performance with spectral predictions than standard practice in regression models. The model makes use of pretraining using quantum chemistry calculations and ensembling of multiple submodels to improve generalizability and performance. The spectral predictions that result are of high quality, showing capability to capture the extreme diversity of spectral forms over chemical space and represent complex peak structures. |
| first_indexed | 2024-09-23T12:10:08Z |
| format | Article |
| id | mit-1721.1/131020 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T12:10:08Z |
| publishDate | 2021 |
| publisher | American Chemical Society (ACS) |
| record_format | dspace |
| spelling | mit-1721.1/1310202022-09-28T00:39:03Z Predicting Infrared Spectra with Message Passing Neural Networks McGill, Charles Forsuelo, Michael Guan, Yanfei Green Jr, William H Massachusetts Institute of Technology. Department of Chemical Engineering McGill, Charles Infrared (IR) spectroscopy remains an important tool for chemical characterization and identification. Chemprop-IR has been developed as a software package for the prediction of IR spectra through the use of machine learning. This work serves the dual purpose of providing a trained general-purpose model for the prediction of IR spectra with ease and providing the Chemprop-IR software framework for the training of new models. In Chemprop-IR, molecules are encoded using a directed message passing neural network, allowing for molecule latent representations to be learned and optimized for the task of spectral predictions. Model training incorporates spectra metrics and normalization techniques that offer better performance with spectral predictions than standard practice in regression models. The model makes use of pretraining using quantum chemistry calculations and ensembling of multiple submodels to improve generalizability and performance. The spectral predictions that result are of high quality, showing capability to capture the extreme diversity of spectral forms over chemical space and represent complex peak structures. DARPA (Contract HR00111920025) 2021-06-17T19:14:21Z 2021-06-17T19:14:21Z 2021-05 Article http://purl.org/eprint/type/JournalArticle 1549-9596 1549-960X https://hdl.handle.net/1721.1/131020 McGill, Charles et al. "Predicting Infrared Spectra with Message Passing Neural Networks." Forthcoming in Journal of Chemical Information and Modeling (2021): doi.org/10.1021/acs.jcim.1c00055. © 2021 American Chemical Society https://doi.org/10.1021/acs.jcim.1c00055 Journal of Chemical Information and Modeling Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf American Chemical Society (ACS) Charles McGill |
| spellingShingle | McGill, Charles Forsuelo, Michael Guan, Yanfei Green Jr, William H Predicting Infrared Spectra with Message Passing Neural Networks |
| title | Predicting Infrared Spectra with Message Passing Neural Networks |
| title_full | Predicting Infrared Spectra with Message Passing Neural Networks |
| title_fullStr | Predicting Infrared Spectra with Message Passing Neural Networks |
| title_full_unstemmed | Predicting Infrared Spectra with Message Passing Neural Networks |
| title_short | Predicting Infrared Spectra with Message Passing Neural Networks |
| title_sort | predicting infrared spectra with message passing neural networks |
| url | https://hdl.handle.net/1721.1/131020 |
| work_keys_str_mv | AT mcgillcharles predictinginfraredspectrawithmessagepassingneuralnetworks AT forsuelomichael predictinginfraredspectrawithmessagepassingneuralnetworks AT guanyanfei predictinginfraredspectrawithmessagepassingneuralnetworks AT greenjrwilliamh predictinginfraredspectrawithmessagepassingneuralnetworks |