{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,5]],"date-time":"2026-05-05T02:41:05Z","timestamp":1777948865168,"version":"3.51.4"},"reference-count":47,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,5,2]],"date-time":"2026-05-02T00:00:00Z","timestamp":1777680000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,5,3]],"date-time":"2026-05-03T00:00:00Z","timestamp":1777766400000},"content-version":"vor","delay-in-days":1,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["J Cheminform"],"abstract":"<jats:sec>\n                    <jats:title>Abstract<\/jats:title>\n                    <jats:p>\n                      Generative models based on diffusion and flow matching have recently been applied to structure-based drug design, but their outputs often include unrealistic protein\u2013ligand interactions that do not obey the laws of physics. We present an energy guidance framework that incorporates a molecular mechanics force field (MMFF94) directly into the sampling process. The method steers molecular generation toward more physically plausible and energetically stable conformations without retraining the underlying model. We evaluate this approach using two state-of-the-art architectures, SemlaFlow, a flow matching model and EDM, a diffusion model, on the PDBBind dataset. Across both models, energy guidance improves enthalpic interaction energy, improves strain energy by up to 75\n                      <jats:inline-formula>\n                        <jats:alternatives>\n                          <jats:tex-math>$$\\%$$<\/jats:tex-math>\n                          <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\">\n                            <mml:mo>%<\/mml:mo>\n                          <\/mml:math>\n                        <\/jats:alternatives>\n                      <\/jats:inline-formula>\n                      , and generates over 1000 ligands with better docking scores than native ligands. These results demonstrate that lightweight, physics-based guidance can significantly enhance generative drug design while preserving chemical validity and diversity.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Scientific contribution<\/jats:title>\n                    <jats:p>\n                      We introduce a novel,\n                      <jats:italic>training-free force field guidance<\/jats:italic>\n                      framework that steers ligand generation using empirical molecular mechanics (e.g., MMFF94) during diffusion or flow-based sampling\u2013without modifying or retraining the base generative model (e.g., EDM or Semflaflow by\u00a0[24]). Our method operates as a plug-in during inference time, leveraging energy feedback to generate poses with lower strain and having better predicted interactions with the protein structure.\n                    <\/jats:p>\n                    <jats:p>\n                      Our main contributions are as follows:\n                      <jats:list list-type=\"bullet\">\n                        <jats:list-item>\n                          <jats:p>Energy-based guidance without retraining: Unlike methods that require gradients from neural affinity predictors (e.g., BADGER\u00a0[26]), our approach injects classical force field feedback (MMFF94) directly during the posterior sampling step.<\/jats:p>\n                        <\/jats:list-item>\n                        <jats:list-item>\n                          <jats:p>Improved docking and strain metrics: In benchmarks against unconditional EDM and Semflaflow, our guided inference yields consistently better AutoDock Vina scores and lower ligand strain energy, even after optimizing the final structures using the same force field.<\/jats:p>\n                        <\/jats:list-item>\n                        <jats:list-item>\n                          <jats:p>Compatibility and flexibility: Because the guidance module is external, it can be applied broadly to multiple generative backbones\u2013without retraining or architecture modifications, and can be applied to arbitrary differentiable potential energy functions.<\/jats:p>\n                        <\/jats:list-item>\n                        <jats:list-item>\n                          <jats:p>Theoretical guarantee of stability. We demonstrate in Appendix B that the gradient correction step corresponds to a descent step on the energy under standard smoothness assumptions. While the full sampling update also includes model-driven (and, in the diffusion case, stochastic) components, this result formalizes how the guidance term locally biases the trajectory toward lower-energy regions and provides a principled justification for its stabilizing effect.<\/jats:p>\n                        <\/jats:list-item>\n                      <\/jats:list>\n                    <\/jats:p>\n                  <\/jats:sec>","DOI":"10.1186\/s13321-026-01198-2","type":"journal-article","created":{"date-parts":[[2026,5,2]],"date-time":"2026-05-02T14:32:26Z","timestamp":1777732346000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Improving protein-ligand complex generation with force field guidance"],"prefix":"10.1186","volume":"18","author":[{"given":"Helen","family":"Lai","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Tingyu","family":"Wang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hassan","family":"Sirelkhatim","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Joe","family":"Eaton","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Howard","family":"Huang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Brad","family":"Rees","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Ola","family":"Engkvist","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jon Paul","family":"Janet","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xiaoyun","family":"Wang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Alessandro","family":"Tibo","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","published-online":{"date-parts":[[2026,5,2]]},"reference":[{"issue":"9","key":"1198_CR1","doi-asserted-by":"publisher","first-page":"2817","DOI":"10.1021\/ja0771033","volume":"130","author":"R Abel","year":"2008","unstructured":"Abel R, Young T, Farid R et al (2008) Role of the active-site solvent in the thermodynamics of factor Xa ligand binding. J Am Chem Soc 130(9):2817\u20132831. https:\/\/doi.org\/10.1021\/ja0771033","journal-title":"J Am Chem Soc"},{"key":"1198_CR2","unstructured":"Albergo MS, Vanden-Eijnden E (2023) Building normalizing flows with stochastic interpolants. In: The eleventh international conference on learning representations"},{"issue":"9","key":"1198_CR3","doi-asserted-by":"publisher","first-page":"787","DOI":"10.1016\/j.chembiol.2003.09.002","volume":"10","author":"AC Anderson","year":"2003","unstructured":"Anderson AC (2003) The process of structure-based drug design. Chem Biol 10(9):787\u2013797. https:\/\/doi.org\/10.1016\/j.chembiol.2003.09.002","journal-title":"Chem Biol"},{"key":"1198_CR4","first-page":"17981","volume":"34","author":"J Austin","year":"2021","unstructured":"Austin J, Johnson DD, Ho J et al (2021) Structured denoising diffusion models in discrete state-spaces. Adv Neural Inf Process Syst 34:17981\u201317993","journal-title":"Adv Neural Inf Process Syst"},{"issue":"1","key":"1198_CR5","doi-asserted-by":"publisher","first-page":"185","DOI":"10.1038\/s41597-022-01288-4","volume":"9","author":"S Axelrod","year":"2022","unstructured":"Axelrod S, G\u00f3mez-Bombarelli R (2022) Geom, energy-annotated molecular conformations for property prediction and molecular generation. Sci Data 9(1):185. https:\/\/doi.org\/10.1038\/s41597-022-01288-4","journal-title":"Sci Data"},{"key":"1198_CR6","unstructured":"Batzolis G, Stanczuk J, Sch\u00f6nlieb CB, et\u00a0al (2021) Conditional image generation with score-based diffusion models. arXiv preprint arXiv:2111.13606"},{"issue":"2","key":"1198_CR7","doi-asserted-by":"publisher","first-page":"90","DOI":"10.1038\/nchem.1243","volume":"4","author":"GR Bickerton","year":"2012","unstructured":"Bickerton GR, Paolini GV, Besnard J et al (2012) Quantifying the chemical beauty of drugs. Nat Chem 4(2):90\u201398","journal-title":"Nat Chem"},{"key":"1198_CR8","unstructured":"Black K, Janner M, Du Y, et\u00a0al (2023) Training diffusion models with reinforcement learning. arXiv preprint arXiv:2305.13301"},{"issue":"1","key":"1198_CR9","doi-asserted-by":"publisher","first-page":"72","DOI":"10.1186\/s13321-021-00548-6","volume":"13","author":"C Bouysset","year":"2021","unstructured":"Bouysset C, Fiorucci S (2021) Prolif: a library to encode molecular interactions as fingerprints. J cheminform 13(1):72","journal-title":"J cheminform"},{"issue":"2","key":"1198_CR10","doi-asserted-by":"publisher","first-page":"187","DOI":"10.1002\/jcc.540040211","volume":"4","author":"BR Brooks","year":"1983","unstructured":"Brooks BR, Bruccoleri RE, Olafson BD et al (1983) Charmm: a program for macromolecular energy, minimization, and dynamics calculations. J Comput Chem 4(2):187\u2013217","journal-title":"J Comput Chem"},{"issue":"9","key":"1198_CR11","doi-asserted-by":"publisher","first-page":"3130","DOI":"10.1039\/D3SC04185A","volume":"15","author":"M Buttenschoen","year":"2024","unstructured":"Buttenschoen M, Morris GM, Deane CM (2024) Posebusters: AI-based docking methods fail to generate physically valid poses or generalise to novel sequences. Chem Sci 15(9):3130\u20133139","journal-title":"Chem Sci"},{"key":"1198_CR12","unstructured":"Campbell A, Yim J, Barzilay R, et\u00a0al (2024) Generative flows on discrete state-spaces: Enabling multimodal flows with applications to protein co-design. In: International conference on machine learning, PMLR, pp 5453\u20135512"},{"issue":"16","key":"1198_CR13","doi-asserted-by":"publisher","first-page":"1668","DOI":"10.1002\/jcc.20290","volume":"26","author":"DA Case","year":"2005","unstructured":"Case DA, Cheatham TE III, Darden T et al (2005) The amber biomolecular simulation programs. J Comput Chem 26(16):1668\u20131688","journal-title":"J Comput Chem"},{"key":"1198_CR14","unstructured":"Corso G, St\u00e4rk H, Jing B, et\u00a0al (2023) Diffdock: Diffusion steps, twists, and turns for molecular docking. In: The eleventh international conference on learning representations"},{"key":"1198_CR15","unstructured":"Cremer J, Irwin R, Tibo A, et\u00a0al (2025) Flowr: flow matching for structure-aware de novo, interaction-and fragment-based ligand generation. arXiv preprint arXiv:2504.10564"},{"key":"1198_CR16","doi-asserted-by":"publisher","DOI":"10.3390\/ijms17020144","author":"X Du","year":"2016","unstructured":"Du X, Li Y, Xia YL et al (2016) Insights into protein-ligand interactions: mechanisms, models, and methods. Int J Mol Sci. https:\/\/doi.org\/10.3390\/ijms17020144","journal-title":"Int J Mol Sci"},{"issue":"7","key":"1198_CR17","doi-asserted-by":"publisher","first-page":"13384","DOI":"10.3390\/molecules200713384","volume":"20","author":"LG Ferreira","year":"2015","unstructured":"Ferreira LG, Dos Santos RN, Oliva G et al (2015) Molecular docking and structure-based drug design strategies. Molecules 20(7):13384\u201313421. https:\/\/doi.org\/10.3390\/molecules200713384","journal-title":"Molecules"},{"issue":"7","key":"1198_CR18","doi-asserted-by":"publisher","first-page":"1739","DOI":"10.1021\/jm0306430","volume":"47","author":"RA Friesner","year":"2004","unstructured":"Friesner RA, Banks JL, Murphy RB et al (2004) Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy. J Med Chem 47(7):1739\u20131749","journal-title":"J Med Chem"},{"issue":"5\u20136","key":"1198_CR19","doi-asserted-by":"publisher","first-page":"490","DOI":"10.1002\/(SICI)1096-987X(199604)17:5\/6<490::AID-JCC1>3.0.CO;2-P","volume":"17","author":"TA Halgren","year":"1996","unstructured":"Halgren TA (1996) Merck molecular force field. i. basis, form, scope, parameterization, and performance of mmff94. J Comput Chem 17(5\u20136):490\u2013519","journal-title":"J Comput Chem"},{"key":"1198_CR20","unstructured":"Harris C, Didi K, Jamasb AR, et\u00a0al (2023) Benchmarking generated poses: how rational is structure-based drug design with generative models? arXiv preprint arXiv:2308.07413"},{"key":"1198_CR21","unstructured":"Ho J, Salimans T (2022) Classifier-free diffusion guidance. arXiv preprint arXiv:2207.12598"},{"key":"1198_CR22","first-page":"6840","volume":"33","author":"J Ho","year":"2020","unstructured":"Ho J, Jain A, Abbeel P (2020) Denoising diffusion probabilistic models. Adv Neural Inf Process Syst 33:6840\u20136851","journal-title":"Adv Neural Inf Process Syst"},{"key":"1198_CR23","unstructured":"Hoogeboom E, Satorras VG, Vignac C, et\u00a0al (2022) Equivariant diffusion for molecule generation in 3d. In: International conference on machine learning, PMLR, pp 8867\u20138887"},{"key":"1198_CR24","unstructured":"Irwin R, Tibo A, Janet JP, et\u00a0al (2025) Semlaflow\u2013efficient 3d molecular generation with latent attention and equivariant flow matching. In: The 28th international conference on artificial intelligence and statistics"},{"key":"1198_CR25","unstructured":"Jain AN, Cleves AE, Walters WP (2024) Deep-learning based docking methods: fair comparisons to conventional docking workflows. arXiv preprint arXiv:2412.02889"},{"key":"1198_CR26","unstructured":"Jian Y, Curtis W, Reidenbach D (2024) General binding affinity guidance for diffusion models in structure-based drug design. In: arxiv"},{"key":"1198_CR27","doi-asserted-by":"publisher","DOI":"10.1016\/j.tet.2020.131865","volume":"79","author":"T Lewis-Atwell","year":"2021","unstructured":"Lewis-Atwell T, Townsend PA, Grayson MN (2021) Comparisons of different force fields in conformational analysis and searching of organic molecules: a review. Tetrahedron 79:131865","journal-title":"Tetrahedron"},{"key":"1198_CR28","unstructured":"Lipman Y, Chen RT, Ben-Hamu H, et\u00a0al (2023) Flow matching for generative modeling. In: The eleventh international conference on learning representations"},{"issue":"24","key":"1198_CR29","doi-asserted-by":"publisher","first-page":"9260","DOI":"10.1021\/acs.jcim.4c01598","volume":"64","author":"H Liu","year":"2024","unstructured":"Liu H, Qin Y, Niu Z et al (2024) How good are current pocket-based 3d generative models?: The benchmark set and evaluation of protein pocket-based 3d molecular generative models. J Chem Inf Model 64(24):9260\u20139275. https:\/\/doi.org\/10.1021\/acs.jcim.4c01598","journal-title":"J Chem Inf Model"},{"key":"1198_CR30","unstructured":"Liu X, Gong C, et\u00a0al (2023) Flow straight and fast: Learning to generate and transfer data with rectified flow. In: The eleventh international conference on learning representations"},{"key":"1198_CR31","doi-asserted-by":"publisher","DOI":"10.1103\/PRXLife.3.023013","volume":"3","author":"Y Liu","year":"2025","unstructured":"Liu Y, Yu Z, Lindsay RJ et al (2025) Exendiff: an experiment-guided diffusion model for protein conformational ensemble generation. PRX Life 3:023013. https:\/\/doi.org\/10.1103\/PRXLife.3.023013","journal-title":"PRX Life"},{"key":"1198_CR32","doi-asserted-by":"crossref","unstructured":"McInnes L, Healy J, Melville J (2018) Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426","DOI":"10.21105\/joss.00861"},{"key":"1198_CR33","unstructured":"Nichol AQ, Dhariwal P (2021) Improved denoising diffusion probabilistic models. In: International conference on machine learning, PMLR, pp 8162\u20138171"},{"key":"1198_CR34","unstructured":"Ramesh A, Dhariwal P, Nichol A, et\u00a0al (2022) Hierarchical text-conditional image generation with clip latents"},{"issue":"25","key":"1198_CR35","doi-asserted-by":"publisher","first-page":"10024","DOI":"10.1021\/ja00051a040","volume":"114","author":"AK Rapp\u00e9","year":"1992","unstructured":"Rapp\u00e9 AK, Casewit CJ, Colwell K et al (1992) Uff, a full periodic table force field for molecular mechanics and molecular dynamics simulations. J Am Chem Soc 114(25):10024\u201310035","journal-title":"J Am Chem Soc"},{"key":"1198_CR36","first-page":"36479","volume":"35","author":"C Saharia","year":"2022","unstructured":"Saharia C, Chan W, Saxena S et al (2022) Photorealistic text-to-image diffusion models with deep language understanding. Adv Neural Inf Process Syst 35:36479\u201336494","journal-title":"Adv Neural Inf Process Syst"},{"key":"1198_CR37","doi-asserted-by":"publisher","DOI":"10.1021\/acs.jcim.5c01020","author":"N Sanjrani","year":"2025","unstructured":"Sanjrani N, Coupry DE, Pog\u00e1ny P et al (2025) Benchmarking 3d structure-based molecule generators. J Chem Inf Model. https:\/\/doi.org\/10.1021\/acs.jcim.5c01020","journal-title":"J Chem Inf Model"},{"key":"1198_CR38","volume-title":"Schr\u00f6dinger Release 2025\u20132: Maestro","author":"Schr\u00f6dinger LLC","year":"2025","unstructured":"Schr\u00f6dinger LLC (2025) Schr\u00f6dinger Release 2025\u20132: Maestro. Schr\u00f6dinger, LLC, New York, NY"},{"key":"1198_CR39","unstructured":"Sohl-Dickstein J, Weiss E, Maheswaranathan N, et\u00a0al (2015) Deep unsupervised learning using nonequilibrium thermodynamics. In: International conference on machine learning, pmlr, pp 2256\u20132265"},{"key":"1198_CR40","unstructured":"Stark H, Jing B, Barzilay R, et\u00a0al (2024) Harmonic self-conditioned flow matching for joint multi-ligand docking and binding site design. In: Forty-first international conference on machine learning"},{"issue":"2","key":"1198_CR41","doi-asserted-by":"publisher","first-page":"455","DOI":"10.1002\/jcc.21334","volume":"31","author":"O Trott","year":"2010","unstructured":"Trott O, Olson AJ (2010) Autodock vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31(2):455\u2013461","journal-title":"J Comput Chem"},{"issue":"12","key":"1198_CR42","doi-asserted-by":"publisher","first-page":"4111","DOI":"10.1021\/jm048957q","volume":"48","author":"R Wang","year":"2005","unstructured":"Wang R, Fang X, Lu Y et al (2005) The pdbbind database: methodologies and updates. J Med Chem 48(12):4111\u20134119","journal-title":"J Med Chem"},{"key":"1198_CR43","unstructured":"Wang Y, Wang L, Shen Y, et\u00a0al (2024) Protein conformation generation via force-guided se (3) diffusion models. arXiv preprint arXiv:2403.14088"},{"key":"1198_CR44","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06415-8","author":"JL Watson","year":"2023","unstructured":"Watson JL, Juergens D, Bennett NR et al (2023) De novo design of protein structure and function with RFdiffusion. Nature. https:\/\/doi.org\/10.1038\/s41586-023-06415-8","journal-title":"Nature"},{"key":"1198_CR45","unstructured":"Zhang S, Zhang W, Gu Q (2025) Energy-weighted flow matching for offline reinforcement learning. arXiv preprint arXiv:2503.04975"},{"key":"1198_CR46","doi-asserted-by":"crossref","unstructured":"Zhang Z, Shen WX, Liu Q, et\u00a0al (2024) Efficient generation of protein pockets with pocketgen. Nature machine intelligence pp 1\u201314","DOI":"10.1101\/2024.02.25.581968"},{"key":"1198_CR47","unstructured":"Zhao Y, Uehara M, Scalia G, et\u00a0al (2024) Adding conditional control to diffusion models with reinforcement learning. arXiv preprint arXiv:2406.12120"}],"container-title":["Journal of Cheminformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/article\/10.1186\/s13321-026-01198-2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s13321-026-01198-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s13321-026-01198-2.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,3]],"date-time":"2026-05-03T21:01:17Z","timestamp":1777842077000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1186\/s13321-026-01198-2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,2]]},"references-count":47,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["1198"],"URL":"https:\/\/doi.org\/10.1186\/s13321-026-01198-2","relation":{},"ISSN":["1758-2946"],"issn-type":[{"value":"1758-2946","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,2]]},"assertion":[{"value":"10 October 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"31 March 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"2 May 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"LLMs were exclusively used for text cleanup purposes.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Use of LLMs"}},{"value":"The authors declare no competing interest.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interest"}}],"article-number":"55"}}