COATI: Multimodal Contrastive Pretraining for Representing and Traversing Chemical Space
Creating a successful small molecule drug is a challenging multiparameter optimization problem in an effectively infinite space of possible molecules. Generative models have emerged as powerful tools for traversing data manifolds composed of images, sounds, and text and offer an opportunity to drama...
Gespeichert in:
Veröffentlicht in: | Journal of chemical information and modeling 2024-02, Vol.64 (4), p.1145-1157 |
---|---|
Hauptverfasser: | , , , , , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | 1157 |
---|---|
container_issue | 4 |
container_start_page | 1145 |
container_title | Journal of chemical information and modeling |
container_volume | 64 |
creator | Kaufman, Benjamin Williams, Edward C. Underkoffler, Carl Pederson, Ryan Mardirossian, Narbe Watson, Ian Parkhill, John |
description | Creating a successful small molecule drug is a challenging multiparameter optimization problem in an effectively infinite space of possible molecules. Generative models have emerged as powerful tools for traversing data manifolds composed of images, sounds, and text and offer an opportunity to dramatically improve the drug discovery and design process. To create generative optimization methods that are more useful than brute-force molecular generation and filtering via virtual screening, we propose that four integrated features are necessary: large, quantitative data sets of molecular structure and activity, an invertible vector representation of realistic accessible molecules, smooth and differentiable regressors that quantify uncertainty, and algorithms to simultaneously optimize properties of interest. Over the course of 12 months, Terray Therapeutics has collected a data set of 2 billion quantitative binding measurements of small molecules to therapeutic targets, which directly motivates multiparameter generative optimization of molecules conditioned on these data. To this end, we present contrastive optimization for accelerated therapeutic inference (COATI), a pretrained, multimodal encoder-decoder model of druglike chemical space. COATI is constructed without any human biasing of features, using contrastive learning from text and 3D representations of molecules to allow for downstream use with structural models. We demonstrate that COATI possesses many of the desired properties of universal molecular embedding: fixed-dimension, invertibility, autoencoding, accurate regression, and low computation cost. Finally, we present a novel metadynamics algorithm for generative optimization using a small subset of our proprietary data collected for a model protein, carbonic anhydrase, designing molecules that satisfy the multiparameter optimization task of potency, solubility, and drug likeness. This work sets the stage for fully integrated generative molecular design and optimization for small molecules. |
doi_str_mv | 10.1021/acs.jcim.3c01753 |
format | Article |
fullrecord | <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_miscellaneous_2922948614</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2922948614</sourcerecordid><originalsourceid>FETCH-LOGICAL-a364t-d4f9cabf291e36a1ce22d15fc0dbfa2737aec3cc1c753e9a196a77ca297b92473</originalsourceid><addsrcrecordid>eNp1kEtLAzEURoMotlb3rmTAjQtb85hJGndl8FGoVLRCd8Nt5o5OmUdNZgr-e1PbuhBc5Sac70tyCDlndMAoZzdg3GBp8nIgDGUqEgeky6JQ97Wk88P9HGnZISfOLSkVQkt-TDpiKJiUMuqSeTwdzca3wVNbNHlZp1AEcV01FlyTrzF4tujnvMqr9yCrbfCCK4sOq2ZzAFUazCys0brNNv7AMje-4HUFBk_JUQaFw7Pd2iNv93ez-LE_mT6M49GkD0KGTT8NM21gkXHNUEhgBjlPWZQZmi4y4EooQCOMYcZ_DzUwLUEpA1yrheahEj1yte1d2fqzRdckZe4MFgVUWLcu4ZpzHQ4lCz16-Qdd1q2t_Os8JXhEtVLcU3RLGVs7ZzFLVjYvwX4ljCYb64m3nmysJzvrPnKxK24XJaa_gb1mD1xvgZ_o_tJ_-74BYE2OWw</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2932509772</pqid></control><display><type>article</type><title>COATI: Multimodal Contrastive Pretraining for Representing and Traversing Chemical Space</title><source>American Chemical Society Journals</source><creator>Kaufman, Benjamin ; Williams, Edward C. ; Underkoffler, Carl ; Pederson, Ryan ; Mardirossian, Narbe ; Watson, Ian ; Parkhill, John</creator><creatorcontrib>Kaufman, Benjamin ; Williams, Edward C. ; Underkoffler, Carl ; Pederson, Ryan ; Mardirossian, Narbe ; Watson, Ian ; Parkhill, John</creatorcontrib><description>Creating a successful small molecule drug is a challenging multiparameter optimization problem in an effectively infinite space of possible molecules. Generative models have emerged as powerful tools for traversing data manifolds composed of images, sounds, and text and offer an opportunity to dramatically improve the drug discovery and design process. To create generative optimization methods that are more useful than brute-force molecular generation and filtering via virtual screening, we propose that four integrated features are necessary: large, quantitative data sets of molecular structure and activity, an invertible vector representation of realistic accessible molecules, smooth and differentiable regressors that quantify uncertainty, and algorithms to simultaneously optimize properties of interest. Over the course of 12 months, Terray Therapeutics has collected a data set of 2 billion quantitative binding measurements of small molecules to therapeutic targets, which directly motivates multiparameter generative optimization of molecules conditioned on these data. To this end, we present contrastive optimization for accelerated therapeutic inference (COATI), a pretrained, multimodal encoder-decoder model of druglike chemical space. COATI is constructed without any human biasing of features, using contrastive learning from text and 3D representations of molecules to allow for downstream use with structural models. We demonstrate that COATI possesses many of the desired properties of universal molecular embedding: fixed-dimension, invertibility, autoencoding, accurate regression, and low computation cost. Finally, we present a novel metadynamics algorithm for generative optimization using a small subset of our proprietary data collected for a model protein, carbonic anhydrase, designing molecules that satisfy the multiparameter optimization task of potency, solubility, and drug likeness. This work sets the stage for fully integrated generative molecular design and optimization for small molecules.</description><identifier>ISSN: 1549-9596</identifier><identifier>ISSN: 1549-960X</identifier><identifier>EISSN: 1549-960X</identifier><identifier>DOI: 10.1021/acs.jcim.3c01753</identifier><identifier>PMID: 38316665</identifier><language>eng</language><publisher>United States: American Chemical Society</publisher><subject>Algorithms ; Carbonic anhydrase ; Datasets ; Design optimization ; Encoders-Decoders ; Machine Learning and Deep Learning ; Molecular structure ; Optimization ; Representations ; Structural models</subject><ispartof>Journal of chemical information and modeling, 2024-02, Vol.64 (4), p.1145-1157</ispartof><rights>2024 American Chemical Society</rights><rights>Copyright American Chemical Society Feb 26, 2024</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-a364t-d4f9cabf291e36a1ce22d15fc0dbfa2737aec3cc1c753e9a196a77ca297b92473</citedby><cites>FETCH-LOGICAL-a364t-d4f9cabf291e36a1ce22d15fc0dbfa2737aec3cc1c753e9a196a77ca297b92473</cites><orcidid>0000-0001-5739-9620</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://pubs.acs.org/doi/pdf/10.1021/acs.jcim.3c01753$$EPDF$$P50$$Gacs$$H</linktopdf><linktohtml>$$Uhttps://pubs.acs.org/doi/10.1021/acs.jcim.3c01753$$EHTML$$P50$$Gacs$$H</linktohtml><link.rule.ids>314,780,784,2765,27076,27924,27925,56738,56788</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/38316665$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Kaufman, Benjamin</creatorcontrib><creatorcontrib>Williams, Edward C.</creatorcontrib><creatorcontrib>Underkoffler, Carl</creatorcontrib><creatorcontrib>Pederson, Ryan</creatorcontrib><creatorcontrib>Mardirossian, Narbe</creatorcontrib><creatorcontrib>Watson, Ian</creatorcontrib><creatorcontrib>Parkhill, John</creatorcontrib><title>COATI: Multimodal Contrastive Pretraining for Representing and Traversing Chemical Space</title><title>Journal of chemical information and modeling</title><addtitle>J. Chem. Inf. Model</addtitle><description>Creating a successful small molecule drug is a challenging multiparameter optimization problem in an effectively infinite space of possible molecules. Generative models have emerged as powerful tools for traversing data manifolds composed of images, sounds, and text and offer an opportunity to dramatically improve the drug discovery and design process. To create generative optimization methods that are more useful than brute-force molecular generation and filtering via virtual screening, we propose that four integrated features are necessary: large, quantitative data sets of molecular structure and activity, an invertible vector representation of realistic accessible molecules, smooth and differentiable regressors that quantify uncertainty, and algorithms to simultaneously optimize properties of interest. Over the course of 12 months, Terray Therapeutics has collected a data set of 2 billion quantitative binding measurements of small molecules to therapeutic targets, which directly motivates multiparameter generative optimization of molecules conditioned on these data. To this end, we present contrastive optimization for accelerated therapeutic inference (COATI), a pretrained, multimodal encoder-decoder model of druglike chemical space. COATI is constructed without any human biasing of features, using contrastive learning from text and 3D representations of molecules to allow for downstream use with structural models. We demonstrate that COATI possesses many of the desired properties of universal molecular embedding: fixed-dimension, invertibility, autoencoding, accurate regression, and low computation cost. Finally, we present a novel metadynamics algorithm for generative optimization using a small subset of our proprietary data collected for a model protein, carbonic anhydrase, designing molecules that satisfy the multiparameter optimization task of potency, solubility, and drug likeness. This work sets the stage for fully integrated generative molecular design and optimization for small molecules.</description><subject>Algorithms</subject><subject>Carbonic anhydrase</subject><subject>Datasets</subject><subject>Design optimization</subject><subject>Encoders-Decoders</subject><subject>Machine Learning and Deep Learning</subject><subject>Molecular structure</subject><subject>Optimization</subject><subject>Representations</subject><subject>Structural models</subject><issn>1549-9596</issn><issn>1549-960X</issn><issn>1549-960X</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><recordid>eNp1kEtLAzEURoMotlb3rmTAjQtb85hJGndl8FGoVLRCd8Nt5o5OmUdNZgr-e1PbuhBc5Sac70tyCDlndMAoZzdg3GBp8nIgDGUqEgeky6JQ97Wk88P9HGnZISfOLSkVQkt-TDpiKJiUMuqSeTwdzca3wVNbNHlZp1AEcV01FlyTrzF4tujnvMqr9yCrbfCCK4sOq2ZzAFUazCys0brNNv7AMje-4HUFBk_JUQaFw7Pd2iNv93ez-LE_mT6M49GkD0KGTT8NM21gkXHNUEhgBjlPWZQZmi4y4EooQCOMYcZ_DzUwLUEpA1yrheahEj1yte1d2fqzRdckZe4MFgVUWLcu4ZpzHQ4lCz16-Qdd1q2t_Os8JXhEtVLcU3RLGVs7ZzFLVjYvwX4ljCYb64m3nmysJzvrPnKxK24XJaa_gb1mD1xvgZ_o_tJ_-74BYE2OWw</recordid><startdate>20240226</startdate><enddate>20240226</enddate><creator>Kaufman, Benjamin</creator><creator>Williams, Edward C.</creator><creator>Underkoffler, Carl</creator><creator>Pederson, Ryan</creator><creator>Mardirossian, Narbe</creator><creator>Watson, Ian</creator><creator>Parkhill, John</creator><general>American Chemical Society</general><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SC</scope><scope>7SR</scope><scope>7U5</scope><scope>8BQ</scope><scope>8FD</scope><scope>JG9</scope><scope>JQ2</scope><scope>L7M</scope><scope>L~C</scope><scope>L~D</scope><scope>7X8</scope><orcidid>https://orcid.org/0000-0001-5739-9620</orcidid></search><sort><creationdate>20240226</creationdate><title>COATI: Multimodal Contrastive Pretraining for Representing and Traversing Chemical Space</title><author>Kaufman, Benjamin ; Williams, Edward C. ; Underkoffler, Carl ; Pederson, Ryan ; Mardirossian, Narbe ; Watson, Ian ; Parkhill, John</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a364t-d4f9cabf291e36a1ce22d15fc0dbfa2737aec3cc1c753e9a196a77ca297b92473</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Algorithms</topic><topic>Carbonic anhydrase</topic><topic>Datasets</topic><topic>Design optimization</topic><topic>Encoders-Decoders</topic><topic>Machine Learning and Deep Learning</topic><topic>Molecular structure</topic><topic>Optimization</topic><topic>Representations</topic><topic>Structural models</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Kaufman, Benjamin</creatorcontrib><creatorcontrib>Williams, Edward C.</creatorcontrib><creatorcontrib>Underkoffler, Carl</creatorcontrib><creatorcontrib>Pederson, Ryan</creatorcontrib><creatorcontrib>Mardirossian, Narbe</creatorcontrib><creatorcontrib>Watson, Ian</creatorcontrib><creatorcontrib>Parkhill, John</creatorcontrib><collection>PubMed</collection><collection>CrossRef</collection><collection>Computer and Information Systems Abstracts</collection><collection>Engineered Materials Abstracts</collection><collection>Solid State and Superconductivity Abstracts</collection><collection>METADEX</collection><collection>Technology Research Database</collection><collection>Materials Research Database</collection><collection>ProQuest Computer Science Collection</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>Computer and Information Systems Abstracts Academic</collection><collection>Computer and Information Systems Abstracts Professional</collection><collection>MEDLINE - Academic</collection><jtitle>Journal of chemical information and modeling</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Kaufman, Benjamin</au><au>Williams, Edward C.</au><au>Underkoffler, Carl</au><au>Pederson, Ryan</au><au>Mardirossian, Narbe</au><au>Watson, Ian</au><au>Parkhill, John</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>COATI: Multimodal Contrastive Pretraining for Representing and Traversing Chemical Space</atitle><jtitle>Journal of chemical information and modeling</jtitle><addtitle>J. Chem. Inf. Model</addtitle><date>2024-02-26</date><risdate>2024</risdate><volume>64</volume><issue>4</issue><spage>1145</spage><epage>1157</epage><pages>1145-1157</pages><issn>1549-9596</issn><issn>1549-960X</issn><eissn>1549-960X</eissn><abstract>Creating a successful small molecule drug is a challenging multiparameter optimization problem in an effectively infinite space of possible molecules. Generative models have emerged as powerful tools for traversing data manifolds composed of images, sounds, and text and offer an opportunity to dramatically improve the drug discovery and design process. To create generative optimization methods that are more useful than brute-force molecular generation and filtering via virtual screening, we propose that four integrated features are necessary: large, quantitative data sets of molecular structure and activity, an invertible vector representation of realistic accessible molecules, smooth and differentiable regressors that quantify uncertainty, and algorithms to simultaneously optimize properties of interest. Over the course of 12 months, Terray Therapeutics has collected a data set of 2 billion quantitative binding measurements of small molecules to therapeutic targets, which directly motivates multiparameter generative optimization of molecules conditioned on these data. To this end, we present contrastive optimization for accelerated therapeutic inference (COATI), a pretrained, multimodal encoder-decoder model of druglike chemical space. COATI is constructed without any human biasing of features, using contrastive learning from text and 3D representations of molecules to allow for downstream use with structural models. We demonstrate that COATI possesses many of the desired properties of universal molecular embedding: fixed-dimension, invertibility, autoencoding, accurate regression, and low computation cost. Finally, we present a novel metadynamics algorithm for generative optimization using a small subset of our proprietary data collected for a model protein, carbonic anhydrase, designing molecules that satisfy the multiparameter optimization task of potency, solubility, and drug likeness. This work sets the stage for fully integrated generative molecular design and optimization for small molecules.</abstract><cop>United States</cop><pub>American Chemical Society</pub><pmid>38316665</pmid><doi>10.1021/acs.jcim.3c01753</doi><tpages>13</tpages><orcidid>https://orcid.org/0000-0001-5739-9620</orcidid></addata></record> |
fulltext | fulltext |
identifier | ISSN: 1549-9596 |
ispartof | Journal of chemical information and modeling, 2024-02, Vol.64 (4), p.1145-1157 |
issn | 1549-9596 1549-960X 1549-960X |
language | eng |
recordid | cdi_proquest_miscellaneous_2922948614 |
source | American Chemical Society Journals |
subjects | Algorithms Carbonic anhydrase Datasets Design optimization Encoders-Decoders Machine Learning and Deep Learning Molecular structure Optimization Representations Structural models |
title | COATI: Multimodal Contrastive Pretraining for Representing and Traversing Chemical Space |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T04%3A06%3A45IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=COATI:%20Multimodal%20Contrastive%20Pretraining%20for%20Representing%20and%20Traversing%20Chemical%20Space&rft.jtitle=Journal%20of%20chemical%20information%20and%20modeling&rft.au=Kaufman,%20Benjamin&rft.date=2024-02-26&rft.volume=64&rft.issue=4&rft.spage=1145&rft.epage=1157&rft.pages=1145-1157&rft.issn=1549-9596&rft.eissn=1549-960X&rft_id=info:doi/10.1021/acs.jcim.3c01753&rft_dat=%3Cproquest_cross%3E2922948614%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2932509772&rft_id=info:pmid/38316665&rfr_iscdi=true |