{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,1]],"date-time":"2026-04-01T21:09:02Z","timestamp":1775077742105,"version":"3.50.1"},"reference-count":59,"publisher":"Oxford University Press (OUP)","license":[{"start":{"date-parts":[[2022,11,18]],"date-time":"2022-11-18T00:00:00Z","timestamp":1668729600000},"content-version":"vor","delay-in-days":321,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2022,11,18]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>The identification of chemical\u2013protein interactions described in the literature is an important task with applications in drug design, precision medicine and biotechnology. Manual extraction of such relationships from the biomedical literature is costly and often prohibitively time-consuming. The BioCreative VII DrugProt shared task provides a benchmark for methods for the automated extraction of chemical\u2013protein relations from scientific text. Here we describe our contribution to the shared task and report on the achieved results. We define the task as a relation classification problem, which we approach with pretrained transformer language models. Upon this basic architecture, we experiment with utilizing textual and embedded side information from knowledge bases as well as additional training data to improve extraction performance. We perform a comprehensive evaluation of the proposed model and the individual extensions including an extensive hyperparameter search leading to 2647 different runs. We find that ensembling and choosing the right pretrained language model are crucial for optimal performance, whereas adding additional data and embedded side information did not improve results. Our best model is based on an ensemble of 10 pretrained transformers and additional textual descriptions of chemicals taken from the Comparative Toxicogenomics Database. The model reaches an F1 score of 79.73% on the hidden DrugProt test set and achieves the first rank out of 107 submitted runs in the official evaluation.<\/jats:p><jats:p>Database URL: https:\/\/github.com\/leonweber\/drugprot<\/jats:p>","DOI":"10.1093\/database\/baac098","type":"journal-article","created":{"date-parts":[[2022,11,18]],"date-time":"2022-11-18T16:39:30Z","timestamp":1668789570000},"source":"Crossref","is-referenced-by-count":14,"title":["Chemical\u2013protein relation extraction with ensembles of carefully tuned pretrained language models"],"prefix":"10.1093","volume":"2022","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-2499-472X","authenticated-orcid":false,"given":"Leon","family":"Weber","sequence":"first","affiliation":[{"name":"Computer Science, Humboldt-Universit\u00e4t zu Berlin , Unter den Linden 6, Berlin 10099, Germany"},{"name":"Group Mathematical Modelling of Cellular Processes, Max Delbr\u00fcck Center for Molecular Medicine in the Helmholtz Association , Robert-R\u00f6ssle-Stra\u00dfe 10, Berlin 13125, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2950-2587","authenticated-orcid":false,"given":"Mario","family":"S\u00e4nger","sequence":"additional","affiliation":[{"name":"Computer Science, Humboldt-Universit\u00e4t zu Berlin , Unter den Linden 6, Berlin 10099, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Samuele","family":"Garda","sequence":"additional","affiliation":[{"name":"Computer Science, Humboldt-Universit\u00e4t zu Berlin , Unter den Linden 6, Berlin 10099, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Fabio","family":"Barth","sequence":"additional","affiliation":[{"name":"Computer Science, Humboldt-Universit\u00e4t zu Berlin , Unter den Linden 6, Berlin 10099, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Christoph","family":"Alt","sequence":"additional","affiliation":[{"name":"Computer Science, Humboldt-Universit\u00e4t zu Berlin , Unter den Linden 6, Berlin 10099, Germany"},{"name":"Research Cluster of Excellence, Science of Intelligence , Marchstr. 23, Berlin 10587, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ulf","family":"Leser","sequence":"additional","affiliation":[{"name":"Computer Science, Humboldt-Universit\u00e4t zu Berlin , Unter den Linden 6, Berlin 10099, Germany"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"286","published-online":{"date-parts":[[2022,11,18]]},"reference":[{"key":"2022111816391692200_R1","doi-asserted-by":"crossref","first-page":"231","DOI":"10.1007\/978-1-4939-9089-4_13","article-title":"Text mining for drug discovery","volume":"1939","author":"Zheng","year":"2019","journal-title":"Methods Mol. Biol. (Clifton, NJ)"},{"key":"2022111816391692200_R2","doi-asserted-by":"crossref","first-page":"183","DOI":"10.1038\/nrd.2017.226","article-title":"Drug development in the era of precision medicine","volume":"17","author":"Dugger","year":"2018","journal-title":"Nat. Rev. Drug Discov."},{"key":"2022111816391692200_R3","doi-asserted-by":"crossref","first-page":"170","DOI":"10.1038\/ng.3774","article-title":"Civic is a community knowledgebase for expert crowdsourcing the clinical interpretation of variants in cancer","volume":"49","author":"Griffith","year":"2017","journal-title":"Nat. Genet."},{"key":"2022111816391692200_R4","first-page":"1","article-title":"Biomedical relation extraction: from binary to complex","volume":"2014","author":"Zhou","year":"2014","journal-title":"Comput. Math. Methods Med."},{"key":"2022111816391692200_R5","article-title":"Exploiting shallow linguistic information for relation extraction from biomedical literature","author":"Giuliano","year":"2006"},{"key":"2022111816391692200_R6","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pcbi.1000837","article-title":"A comprehensive benchmark of kernel methods to extract protein\u2013protein interactions from literature","volume":"6","author":"Tikk","year":"2010","journal-title":"PLoS Comput. Biol."},{"key":"2022111816391692200_R7","doi-asserted-by":"crossref","first-page":"i490","DOI":"10.1093\/bioinformatics\/btaa430","article-title":"PEDL: extracting protein\u2013protein associations using deep language models and distant supervision","volume":"36","author":"Weber","year":"2020","journal-title":"Bioinformatics"},{"key":"2022111816391692200_R8","doi-asserted-by":"crossref","first-page":"83","DOI":"10.1016\/j.jbi.2018.03.011","article-title":"A hybrid model based on neural networks for biomedical relation extraction","volume":"81","author":"Zhang","year":"2018","journal-title":"J. Biomed. Inf."},{"key":"2022111816391692200_R9","first-page":"pp. 1388","article-title":"Fine-tuning pre-trained transformer language models to distantly supervised relation extraction","author":"Alt","year":"2019"},{"key":"2022111816391692200_R10","doi-asserted-by":"crossref","first-page":"1234","DOI":"10.1093\/bioinformatics\/btz682","article-title":"BioBERT: a pre-trained biomedical language representation model for biomedical text mining","volume":"36","author":"Lee","year":"2020","journal-title":"Bioinformatics"},{"key":"2022111816391692200_R11","doi-asserted-by":"crossref","first-page":"2792","DOI":"10.1093\/bioinformatics\/btab042","article-title":"HunFlair: an easy-to-use tool for state-of-the-art biomedical named entity recognition","volume":"37","author":"Weber","year":"2021","journal-title":"Bioinformatics"},{"key":"2022111816391692200_R12","article-title":"Pre-trained language model for biomedical question answering","author":"Yoon","year":"2019"},{"key":"2022111816391692200_R13","first-page":"1","article-title":"Domain-specific language model pretraining for biomedical natural language processing","volume":"3","author":"Yu","year":"2021","journal-title":"ACM Trans. Comput. Healthcare (HEALTH)"},{"key":"2022111816391692200_R14","first-page":"pp. 1107","article-title":"Very deep convolutional networks for text classification","author":"Conneau","year":"2017"},{"key":"2022111816391692200_R15","volume-title":"COLING","author":"Dai","year":"2020"},{"key":"2022111816391692200_R16","first-page":"pp. 6382","article-title":"Eda: Easy data augmentation techniques for boosting performance on text classification tasks","author":"Wei","year":"2019"},{"key":"2022111816391692200_R17","first-page":"pp. 5303","article-title":"Unsupervised paraphrasing consistency training for low resource named entity recognition","author":"Wang","year":"2021"},{"key":"2022111816391692200_R18","first-page":"pp. 856","article-title":"SwitchOut: an efficient data augmentation algorithm for neural machine translation","author":"Wang","year":"2018"},{"key":"2022111816391692200_R19","first-page":"pp. 452","article-title":"Contextual augmentation: data augmentation by words with paradigmatic relations","author":"Kobayashi","year":"2018"},{"key":"2022111816391692200_R20","first-page":"pp. 1257","article-title":"Reside: improving distantly-supervised neural relation extraction using side information","author":"Vashishth","year":"2018"},{"key":"2022111816391692200_R21","first-page":"pp. 3201","article-title":"Connecting language and knowledge with heterogeneous representations for neural relation extraction","author":"Peng","year":"2019"},{"key":"2022111816391692200_R22","article-title":"CoCoScore: context-aware co-occurrence scoring for text mining applications using distant supervision","volume":"06","author":"Junge","year":"2019","journal-title":"Bioinformatics"},{"key":"2022111816391692200_R23","first-page":"pp. 77","article-title":"Constructing Biological Knowledge Bases by Extracting Information from Text Sources","author":"Craven","year":"1999"},{"key":"2022111816391692200_R24","first-page":"pp. 120","article-title":"Distant Supervision for Cancer Pathway Extraction from Text","author":"Poon","year":"2015"},{"key":"2022111816391692200_R25","first-page":"pp. 1171","article-title":"Distant Supervision for Relation Extraction beyond the Sentence Boundary","author":"Quirk","year":"2017"},{"key":"2022111816391692200_R26","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1186\/s12859-015-0549-5","article-title":"Knowlife: a versatile approach for constructing a large knowledge graph for biomedical sciences","volume":"16","author":"Ernst","year":"2015","journal-title":"BMC Bioinf."},{"key":"2022111816391692200_R27","first-page":"pp. 1003","article-title":"Distant supervision for relation extraction without labeled data","author":"Mintz","year":"2009"},{"key":"2022111816391692200_R28","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pcbi.1005017","article-title":"Text mining genotype-phenotype relationships from biomedical literature for database curation and precision medicine","volume":"12","author":"Singhal","year":"2016","journal-title":"PLoS Comput. Biol."},{"key":"2022111816391692200_R29","first-page":"pp. 141","article-title":"Overview of the BioCreative VI chemical-protein interaction track","author":"Krallinger","year":"2017"},{"key":"2022111816391692200_R30","first-page":"pp. 11","article-title":"Overview of DrugProt BioCreative VII track: quality evaluation and large scale text mining of drug-gene\/protein relations","author":"Miranda","year":"2021"},{"key":"2022111816391692200_R31","doi-asserted-by":"crossref","first-page":"D1138","DOI":"10.1093\/nar\/gkaa891","article-title":"Comparative toxicogenomics database (CTD): update 2021","volume":"49","author":"Davis","year":"2021","journal-title":"Nucleic Acids Res."},{"key":"2022111816391692200_R32","doi-asserted-by":"crossref","first-page":"D36","DOI":"10.1093\/nar\/gku1055","article-title":"Gene: a gene-centered information resource at NCBI","volume":"43","author":"Brown","year":"2015","journal-title":"Nucleic Acids Res."},{"key":"2022111816391692200_R33","first-page":"pp. 3641","article-title":"Biomedical entity representations with synonym marginalization","author":"Sung","year":"2020"},{"key":"2022111816391692200_R34","first-page":"1","article-title":"BioCreative V CDR task corpus: a resource for chemical disease relation extraction","volume":"2016","author":"Jiao","year":"2016","journal-title":"Database"},{"key":"2022111816391692200_R35","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1186\/gb-2008-9-s2-s3","article-title":"Overview of BioCreative II gene normalization","volume":"9","author":"Morgan","year":"2008","journal-title":"Genome Biol."},{"key":"2022111816391692200_R36","first-page":"pp. 6710","article-title":"Fair evaluation in concept normalization: a large-scale comparative analysis for BERT-based models","author":"Tutubalina","year":"2020"},{"key":"2022111816391692200_R37","article-title":"Adam: a method for stochastic optimization","author":"Kingma","year":"2015"},{"key":"2022111816391692200_R38","first-page":"pp. 8342","article-title":"Don\u2019t stop pretraining: adapt language models to domains and tasks","author":"Gururangan","year":"2020"},{"key":"2022111816391692200_R39","first-page":"pp. 146","article-title":"Pretrained language models for biomedical and clinical tasks: understanding and extending the state-of-the-art","author":"Lewis","year":"2020"},{"key":"2022111816391692200_R40","doi-asserted-by":"crossref","first-page":"D1074","DOI":"10.1093\/nar\/gkx1037","article-title":"DrugBank 5.0: a major update to the DrugBank database for 2018","volume":"46","author":"Wishart","year":"2018","journal-title":"Nucleic Acids Res."},{"key":"2022111816391692200_R41","doi-asserted-by":"crossref","first-page":"D480","DOI":"10.1093\/nar\/gkaa1100","article-title":"UniProt: the universal protein knowledgebase in 2021","volume":"49","author":"Consortium","year":"2021","journal-title":"Nucleic Acids Res."},{"key":"2022111816391692200_R42","first-page":"1","article-title":"PyKEEN 1.0: A Python Library for Training and Evaluating Knowledge Graph Embeddings","volume":"22","author":"Ali","year":"2021","journal-title":"J. Mach. Learn Res."},{"key":"2022111816391692200_R43","first-page":"pp. 739","article-title":"DGL-KE: training knowledge graph embeddings at scale","author":"Xiang","year":"2020"},{"key":"2022111816391692200_R44","first-page":"pp. 807","article-title":"Rectified linear units improve restricted Boltzmann machines","author":"Nair","year":"2010"},{"key":"2022111816391692200_R45","first-page":"1929","article-title":"Dropout: a simple way to prevent neural networks from overfitting","volume":"15","author":"Srivastava","year":"2014","journal-title":"The Journal of Machine Learning research"},{"key":"2022111816391692200_R46","doi-asserted-by":"crossref","first-page":"236","DOI":"10.1093\/bioinformatics\/btaa674","article-title":"Large-scale entity representation learning for biomedical relationship extraction","volume":"37","author":"S\u00e4nger","year":"2021","journal-title":"Bioinformatics"},{"key":"2022111816391692200_R47","first-page":"pp. 1","article-title":"Embedding entities and relations for learning and inference in knowledge bases","author":"Yang","year":"2015"},{"key":"2022111816391692200_R48","first-page":"pp. 2071","article-title":"Complex embeddings for simple link prediction","author":"Trouillon","year":"2016"},{"key":"2022111816391692200_R49","first-page":"1","author":"Krompa\u00df","year":"2013"},{"key":"2022111816391692200_R50","article-title":"Facebook FAIR\u2019s WMT19 news translation task submission","author":"Nathan","year":"2019"},{"key":"2022111816391692200_R51","first-page":"pp. 479","article-title":"OPUS-MT\u2013building open translation services for the world","author":"Tiedemann","year":"2020"},{"key":"2022111816391692200_R52","article-title":"Roberta: A robustly optimized bert pretraining approach","author":"Liu","year":"2019"},{"key":"2022111816391692200_R53","first-page":"pp. 754","article-title":"An efficient approach for assessing hyperparameter importance","author":"Hutter","year":"2014"},{"key":"2022111816391692200_R54","doi-asserted-by":"crossref","first-page":"D1398","DOI":"10.1093\/nar\/gkab953","article-title":"Therapeutic target database update 2022: facilitating drug discovery with enriched comparative data of targeted agents","volume":"50","author":"Zhou","year":"2022","journal-title":"Nucleic Acids Res."},{"key":"2022111816391692200_R55","doi-asserted-by":"crossref","first-page":"W587","DOI":"10.1093\/nar\/gkz389","article-title":"PubTator central: automated concept annotation for biomedical full text articles","volume":"47","author":"Wei","year":"2019","journal-title":"Nucleic Acids Res."},{"key":"2022111816391692200_R56","first-page":"pp. 26","article-title":"Extracting drug-protein interaction using an ensemble of biomedical pre-trained language models through sequence labeling and text classification techniques","author":"Luo","year":"2021"},{"key":"2022111816391692200_R57","first-page":"pp. 31","article-title":"Using knowledge base to refine data augmentation for biomedical relation extraction","author":"Yoon","year":"2021"},{"key":"2022111816391692200_R58","first-page":"pp. 36","article-title":"UTHealth@BioCreativeVII: domain-specific transformer models for drug-protein relation extraction","author":"Das","year":"2021"},{"key":"2022111816391692200_R59","first-page":"pp. 221","article-title":"BioM-transformers: building large biomedical language models with BERT, ALBERT and ELECTRA","author":"Alrowili","year":"2021"}],"container-title":["Database"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/database\/article-pdf\/doi\/10.1093\/database\/baac098\/47131901\/baac098.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/database\/article-pdf\/doi\/10.1093\/database\/baac098\/47131901\/baac098.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,3,9]],"date-time":"2023-03-09T14:02:18Z","timestamp":1678370538000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/database\/article\/doi\/10.1093\/database\/baac098\/6833204"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2022,1,1]]},"references-count":59,"URL":"https:\/\/doi.org\/10.1093\/database\/baac098","relation":{},"ISSN":["1758-0463"],"issn-type":[{"value":"1758-0463","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2022,1,1]]},"published":{"date-parts":[[2022,1,1]]},"article-number":"baac098"}}