{"id":12287,"date":"2025-02-12T00:50:21","date_gmt":"2025-02-12T00:50:21","guid":{"rendered":"https:\/\/www.blopig.com\/blog\/?p=12287"},"modified":"2025-02-12T00:50:23","modified_gmt":"2025-02-12T00:50:23","slug":"narrowing-the-gap-between-machine-learning-scoring-functions-and-free-energy-perturbation-using-augmented-data","status":"publish","type":"post","link":"https:\/\/www.blopig.com\/blog\/2025\/02\/narrowing-the-gap-between-machine-learning-scoring-functions-and-free-energy-perturbation-using-augmented-data\/","title":{"rendered":"Narrowing the gap between machine learning scoring functions and free energy perturbation using augmented data"},"content":{"rendered":"\n

I’m delighted to report our collaboration (<\/a>\u00cdsak Valsson<\/a>, <\/a>Matthew Warren<\/a>, <\/a>Aniket Magarkar<\/a>, <\/a>Phil Biggin<\/a>, & <\/a>Charlotte Deane<\/a>), on “Narrowing the gap between machine learning scoring functions and free energy perturbation using augmented data”, has been published in Nature’s Communications Chemistry<\/em> (https:\/\/doi.org\/10.1038\/s42004-025-01428-y<\/a>).<\/p>\n\n\n\n

\"\"<\/a><\/figure>\n\n\n\n


During his MSc dissertation project in the Department of Statistics, University of Oxford<\/a>, OPIG member <\/a>\u00cdsak Valsson<\/a> developed an attention-based GNN to predict protein-ligand binding affinity called “AEV-PLIG”. It featurizes a ligand’s atoms using Atomic Environment Vectors to describe the Protein-Ligand Interactions found in a 3D protein-ligand complex. AEV-PLIG is free and open source (BSD 3-Clause), available from GitHub at https:\/\/github.com\/oxpig\/AEV-PLIG<\/a>, and forked at https:\/\/github.com\/bigginlab\/AEV-PLIG<\/a>.<\/p>\n\n\n\n\n\n\n\n

\u00cdsak also developed a much more challenging protein-ligand binding affinity prediction benchmark than CASF-2016, called “Out-Of-Distribution Test”, which is also available (https:\/\/github.com\/isakvals\/OOD-Test<\/a>). It is designed to assess how well a method generalizes to more dissimilar ligands and proteins than seen its training set. AEV-PLIG performed best in terms of Pearson correlation coefficient on OOD Test, and another tough benchmark also developed in OPIG called “0-Ligand Bias” (https:\/\/doi.org\/10.1093\/bioinformatics\/btaf040<\/a>). AEV-PLIG proved to be more accurate than RF-score, Pafnucy, OnionNet-2, PointVS, SIGN, and AEScore (Table 1).

<\/a>Matthew Warren<\/a> showed that augmenting our training data (PDBbind v2020) with semi-synthetic data (BindingNet) boosted the performance of AEScore (https:\/\/github.com\/RMeli\/aescore<\/a>) and we confirmed this with AEV-PLIG.

Together, we found that with even more data (BindingDB), our augmented AEV-PLIG model’s prediction accuracy starts to approach that of Free Energy Perturbation (FEP+) for congeneric series of ligands that bind the same protein \u2014 see Figures 3 & 4 in our paper \u2014 yet we are ~400,000 times faster, while using a single GPU instead of several. We also showed AEV-PLIG performed best on our FEP Benchmark than the other ML-based scoring functions we looked at (Table 1).

Another take home: the performance of AEV-PLIG steadily improved as we increased the fraction of augmented training data with no sign of leveling off (Figure S5). Whether this putative ‘scaling law’ applies to protein-ligand binding affinity prediction remains to be seen, but the notion of a simpler model architecture with more physically-relevant features combined with more data shows great promise…<\/p>\n","protected":false},"excerpt":{"rendered":"

I’m delighted to report our collaboration (\u00cdsak Valsson, Matthew Warren, Aniket Magarkar, Phil Biggin, & Charlotte Deane), on “Narrowing the gap between machine learning scoring functions and free energy perturbation using augmented data”, has been published in Nature’s Communications Chemistry (https:\/\/doi.org\/10.1038\/s42004-025-01428-y). During his MSc dissertation project in the Department of Statistics, University of Oxford, OPIG […]<\/p>\n","protected":false},"author":35,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"nf_dc_page":"","wikipediapreview_detectlinks":true,"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"ngg_post_thumbnail":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[633,187,29,361,189,274,291,202,227,201],"tags":[837,835,832,455,836,831,288,834],"ppma_author":[488],"class_list":["post-12287","post","type-post","status-publish","format-standard","hentry","category-ai","category-cheminformatics","category-code","category-data-science","category-machine-learning","category-molecular-recognition","category-protein-ligand-docking","category-proteins","category-python-code","category-small-molecules","tag-acsfs","tag-augmented-data","tag-fep","tag-gnns","tag-physicochemical-deep-learning","tag-protein-ligand-binding","tag-scoring-functions","tag-synthetic-data"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"authors":[{"term_id":488,"user_id":35,"is_guest":0,"slug":"garrett","display_name":"Garrett","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/df625261419c37dd5c5937e37f17a732626acd6eea1e6fabd03d935c25b453bf?s=96&d=mm&r=g","0":null,"1":"","2":"","3":"","4":"","5":"","6":"","7":"","8":""}],"_links":{"self":[{"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/posts\/12287","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/users\/35"}],"replies":[{"embeddable":true,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/comments?post=12287"}],"version-history":[{"count":1,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/posts\/12287\/revisions"}],"predecessor-version":[{"id":12289,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/posts\/12287\/revisions\/12289"}],"wp:attachment":[{"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/media?parent=12287"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/categories?post=12287"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/tags?post=12287"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.blopig.com\/blog\/wp-json\/wp\/v2\/ppma_author?post=12287"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}