{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/message-passing-neural-networks-for-high","title":"Message-passing neural networks for high-throughput polymer screening","arxiv_id":"1807.10363","date":"2018-07-26","proceeding":null,"authors":["Peter C. St. John","Caleb Phillips","Travis W. Kemper","A. Nolan Wilson","Michael F. Crowley","Mark R. Nimlos","Ross E. Larsen"],"abstract":"Machine learning methods have shown promise in predicting molecular\nproperties, and given sufficient training data machine learning approaches can\nenable rapid high-throughput virtual screening of large libraries of compounds.\nGraph-based neural network architectures have emerged in recent years as the\nmost successful approach for predictions based on molecular structure, and have\nconsistently achieved the best performance on benchmark quantum chemical\ndatasets. However, these models have typically required optimized 3D structural\ninformation for the molecule to achieve the highest accuracy. These 3D\ngeometries are costly to compute for high levels of theory, limiting the\napplicability and practicality of machine learning methods in high-throughput\nscreening applications. In this study, we present a new database of candidate\nmolecules for organic photovoltaic applications, comprising approximately\n91,000 unique chemical structures.Compared to existing datasets, this dataset\ncontains substantially larger molecules (up to 200 atoms) as well as\nextrapolated properties for long polymer chains. We show that message-passing\nneural networks trained with and without 3D structural information for these\nmolecules achieve similar accuracy, comparable to state-of-the-art methods on\nexisting benchmark datasets. These results therefore emphasize that for larger\nmolecules with practical applications, near-optimal prediction results can be\nobtained without using optimized 3D geometry as an input. We further show that\nlearned molecular representations can be leveraged to reduce the training data\nrequired to transfer predictions to a new DFT functional.","url_abs":"http://arxiv.org/abs/1807.10363v2","url_pdf":"http://arxiv.org/pdf/1807.10363v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"message-passing-neural-networks-for-high","repo_url":"https://github.com/nrel/nfp","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"3d-geometry","task_name":"3D geometry"},{"task_slug":"machine-learning","task_name":"BIG-bench Machine Learning"},{"task_slug":"high","task_name":"Vocal Bursts Intensity Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1807.10363","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}