Physics-informed deep learning: A new paradigm for macromolecular recognition and rational protein design
-
GRAPHICAL ABSTRACT
DownLoad:
Full size image
-
PUBLIC SUMMARY
-
PIDL grounds biomolecular AI in physical laws to reduce implausible structural predictions.
Three pillars connect physical constraints, equivariant architectures, and guided representations.
Physics-aware models improve recognition across ligands, peptides, proteins, and immune complexes.
Future systems should unite dynamics, cellular context, and closed-loop experimental validation.
-
-
ABSTRACT
Targeted drug discovery is fundamentally bottlenecked by the challenge of accurately modeling complex biomolecular interactions, ranging from small-molecule ligand binding to high-order macromolecular assemblies. While traditional physics-based computational methods provide profound mechanistic insights, their clinical utility is frequently hampered by prohibitive computational costs and scalability limitations when addressing highly flexible, cross-scale systems. Conversely, the rapid emergence of pure deep learning offers unprecedented computational speed but suffers from a fundamental “black-box” nature, sometimes yielding physically improbable conformations—often referred to as “hallucinations”—that can pose challenges in real-world experimental validation. To bridge this critical translational gap, the integration of physical principles with artificial intelligence—Physics-Informed Deep Learning (PIDL)—is currently driving a fundamental transition from purely empirical approximations to rational, physically grounded design. This review constructs a strategic framework to critically evaluate these transformative advances, structured around three methodological pillars: (1) Physics-constrained optimization, which integrates thermodynamic principles and integrative experimental restraints at the output level to decode macromolecular dynamics; (2) Physics-encoded architectures, which embed appropriate SE(3) or E(3) geometric symmetries directly into neural network topologies for precise structural recognition; and (3) Physics-guided representations, which project discrete sequences into continuous physicochemical manifolds to enhance interaction prediction. By delineating how physics-based priors synergize with data-driven representation learning, this review not only synthesizes current algorithmic breakthroughs but also provides a comprehensive roadmap for generating physically plausible and thermodynamically stable therapeutics, ultimately accelerating the transition of computationally designed molecules from in silico blueprints to viable clinical candidates. -
-
REFERENCES
[1] Bohacek R. S., McMartin C. and Guida W. C. (1996). The art and practice of structure-based drug design: a molecular modeling perspective. Med. Res. Rev. 16:3−50. DOI:3.0.CO;2-6">10.1002/(SICI)1098-1128(199601)16:1<3::AID-MED1>3.0.CO;2-6 [2] Reymond J.-L. and Awale M. (2012). Exploring chemical space for drug discovery using the chemical universe database. ACS Chem. Neurosci. 3:649−657. DOI:10.1021/cn3000422 [3] Sun D., Gao W., Hu H. et al. (2022). Why 90% of clinical drug development fails and how to improve it?. Acta Pharm. Sin. B 12:3049−3062. DOI:10.1016/j.apsb.2022.02.002 [4] Oprea T. I., Bologa C. G., Brunak S. et al. (2018). Unexplored therapeutic opportunities in the human genome. Nat. Rev. Drug Discov. 17:317−332. DOI:10.1038/nrd.2018.14 [5] Lee C. H., Huh J., Buckley P. R. et al. (2023). A robust deep learning workflow to predict CD8+ T-cell epitopes. Genome Med. 15:70. DOI:10.1186/s13073-023-01225-z [6] De Vivo M., Masetti M., Bottegoni G. et al. (2016). Role of molecular dynamics and related methods in drug discovery. J. Med. Chem. 59:4035−4061. DOI:10.1021/acs.jmedchem.5b01684 [7] Claveria-Gimeno R., Vega S., Abian O. et al. (2017). A look at ligand binding thermodynamics in drug discovery. Expert Opin. Drug Discov. 12:363−377. DOI:10.1080/17460441.2017.1297418 [8] Teague S. J. (2003). Implications of protein flexibility for drug discovery. Nat. Rev. Drug Discov. 2:527−541. DOI:10.1038/nrd1129 [9] Lu T., Zhang Z., Zhu J. et al. (2021). Deep learning-based prediction of the T cell receptor–antigen binding specificity. Nat. Mach. Intell. 3:864−875. DOI:10.1038/s42256-021-00383-2 [10] DiMasi J. A., Grabowski H. G. and Hansen R. W. (2016). Innovation in the pharmaceutical industry: new estimates of R&D costs. J. Health Econ. 47:20−33. DOI:10.1016/j.jhealeco.2016.01.012 [11] Arstila T. P., Casrouge A., Baron V. et al. (1999). A direct estimate of the human αβ T cell receptor diversity. Science 286:958−961. DOI:10.1126/science.286.5441.958 [12] Dudek A. Z., Arodz T. and Gálvez J. (2006). Computational methods in developing quantitative structure-activity relationships (QSAR): a review. Comb. Chem. High Throughput Screen. 9:213−228. DOI:10.2174/138620706776055539 [13] Nantasenamat C. (2009). A practical overview of quantitative structure activity relationship. EXCLI J. 8:74−88. DOI:10.17877/DE290R-690 [14] Cherkasov A., Muratov E. N., Fourches D. et al. (2014). QSAR modeling: where have you been? Where are you going to?. J. Med. Chem. 57:4977−5010. DOI:10.1021/jm4004285 [15] Keifer P. A. (2003). Flow injection analysis NMR (FIA–NMR): a novel flow NMR technique that complements LC–NMR and direct injection NMR (DI–NMR). Magn. Reson. Chem. 41:509−516. DOI:10.1002/mrc.1206 [16] Morris C. J. and Della Corte D. (2021). Using molecular docking and molecular dynamics to investigate protein–ligand interactions. Mod. Phys. Lett. B 35:2130002. DOI:10.1142/S0217984921300027 [17] Hollingsworth S. A. and Dror R. O. (2018). Molecular dynamics simulation for all. Neuron 99:1129−1143. DOI:10.1016/j.neuron.2018.08.011 [18] Morris G. M. and Lim-Wilby M. (2008). Molecular docking. Methods Mol. Biol. 443:365−382. DOI:10.1007/978-1-59745-177-2_19 [19] Li J., Fu A. and Zhang L. (2019). An overview of scoring functions used for protein–ligand interactions in molecular docking. Interdiscip. Sci. 11:320−328. DOI:10.1007/s12539-019-00327-w [20] Trott O. and Olson A. J. (2010). AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J. Comput. Chem. 31:455−461. DOI:10.1002/jcc.21334 [21] Wang L., Wu Y., Deng Y. et al. (2015). Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field. J. Am. Chem. Soc. 137:2695−2703. DOI:10.1021/ja512751q [22] Gilson M. K. and Zhou H.-X. (2007). Calculation of protein-ligand binding affinities. Annu. Rev. Biophys. Biomol. Struct. 36:21−42. DOI:10.1146/annurev.biophys.36.040306.132550 [23] Dror R. O., Dirks R. M., Grossman J. P. et al. (2012). Biomolecular simulation: a computational microscope for molecular biology. Annu. Rev. Biophys. 41:429−452. DOI:10.1146/annurev-biophys-042910-155245 [24] Perez J. J., Perez R. A. and Perez A. (2021). Computational modeling as a tool to investigate PPI: from drug design to tissue engineering. Front. Mol. Biosci. 8:681617. DOI:10.3389/fmolb.2021.681617 [25] Sousa S. F., Fernandes P. A. and Ramos M. J. (2006). Protein–ligand docking: current status and future challenges. Proteins 65:15−26. DOI:10.1002/prot.21082 [26] Mohan V., Gibbs A. C., Cummings M. D. et al. (2005). Docking: successes and challenges. Curr. Pharm. Des. 11:323−333. DOI:10.2174/1381612053382106 [27] Yang K., Swanson K., Jin W. et al. (2019). Analyzing learned molecular representations for property prediction. J. Chem. Inf. Model. 59:3370−3388. DOI:10.1021/acs.jcim.9b00237 [28] Altae-Tran H., Ramsundar B., Pappu A. S. et al. (2017). Low data drug discovery with one-shot learning. ACS Cent. Sci. 3:283−293. DOI:10.1021/acscentsci.6b00367 [29] Ching T., Himmelstein D. S., Beaulieu-Jones B. K. et al. (2018). Opportunities and obstacles for deep learning in biology and medicine. J. R. Soc. Interface 15:20170387. DOI:10.1098/rsif.2017.0387 [30] Rossjohn J., Gras S., Miles J. J. et al. (2015). T cell antigen receptor recognition of antigen-presenting molecules. Annu. Rev. Immunol. 33:169−200. DOI:10.1146/annurev-immunol-032414-112334 [31] Knapp B., Demharter S., Esmaielbeiki R. et al. (2015). Current status and future challenges in T-cell receptor/peptide/MHC molecular dynamics simulations. Brief. Bioinform. 16:1035−1044. DOI:10.1093/bib/bbv005 [32] Chmiela S. et al. (2017). Machine learning of accurate energy-conserving molecular force fields. Sci. Adv. 3:e1603015. DOI:10.1126/sciadv.1603015 [33] Unke O. T. et al. (2021). Machine learning force fields. Chem. Rev. 121:10142−10186. DOI:10.1021/acs.chemrev.0c01111 [34] Jiménez-Luna J., Grisoni F. and Schneider G. (2020). Drug discovery with explainable artificial intelligence. Nat. Mach. Intell. 2:573−584. DOI:10.1038/s42256-020-00236-4 [35] Karniadakis G. E. et al. (2021). Physics-informed machine learning. Nat. Rev. Phys. 3:422−440. DOI:10.1038/s42254-021-00314-5 [36] Miyazawa S. and Jernigan R. L. (1996). Residue–residue potentials with a favorable contact pair term and an unfavorable high packing density term. J. Mol. Biol. 256:623−644. DOI:10.1006/jmbi.1996.0114 [37] Wang Y., Li Y., Chen J. et al. (2025). Modeling protein–ligand interactions for drug discovery in the era of deep learning. Chem. Soc. Rev. 54. DOI:10.1039/d5cs00415b DOI:10.1039/d5cs00415b [38] Lei Y. et al. (2021). A deep-learning framework for multi-level peptide–protein interaction prediction. Nat. Commun. 12:5465. DOI:10.1038/s41467-021-25772-4 [39] Moon S., Zhung W., Yang S. et al. (2022). PIGNet: a physics-informed deep learning model toward generalized drug–target interaction predictions. Chem. Sci. 13:3661−3673. DOI:10.1039/d1sc06946b [40] Deshpande A. (2025). Learning biomolecular motion: the physics-informed machine learning paradigm. arXiv: 2511.06585. [41] Ugurlu S. Y. et al. (2024). Cobdock: an accurate and practical machine learning-based consensus blind docking method. J. Cheminform. 16:5. DOI:10.1186/s13321-023-00793-x [42] Honorato R. V. et al. (2024). The HADDOCK2.4 web server for integrative modeling of biomolecular complexes. Nat. Protoc. 19:33. DOI:10.1038/s41596-024-01011-0 [43] Dominguez C., Boelens R. and Bonvin A. M. (2003). HADDOCK: a protein–protein docking approach based on biochemical or biophysical information. J. Am. Chem. Soc. 125:1731−1737. DOI:10.1021/ja026939x [44] Koukos P. I., Réau M. and Bonvin A. M. (2021). Shape-restrained modeling of protein–small-molecule complexes with high ambiguity driven DOCKing. J. Chem. Inf. Model. 61:4807−4818. DOI:10.1021/acs.jcim.1c00796 [45] Nilges M. (1995). Calculation of protein structures with ambiguous distance restraints: automated assignment of ambiguous NOE crosspeaks and disulphide connectivities. J. Mol. Biol. 245:645−660. DOI:10.1006/jmbi.1994.0053 [46] Rout M. P. and Sali A. (2019). Principles for integrative structural biology studies. Cell 177:1384−1403. DOI:10.1016/j.cell.2019.05.016 [47] Vajda S. and Kozakov D. (2009). Convergence and combination of methods in protein–protein docking. Curr. Opin. Struct. Biol. 19:164−170. DOI:10.1016/j.sbi.2009.02.008 [48] Rodrigues J. P. and Bonvin A. M. (2014). Integrative computational modeling of protein interactions. FEBS J. 281:1988−2003. DOI:10.1111/febs.12771 [49] Bonvin A. M. (2006). Flexible protein–protein docking. Curr. Opin. Struct. Biol. 16:194−200. DOI:10.1016/j.sbi.2006.02.002 [50] Rodrigues J. P. G. L. M., Karaca E. and Bonvin A. M. J. J. (2014). Information-driven structural modelling of protein–protein interactions. In: Kukol A. (ed.), Molecular Modeling of Proteins (Springer), pp. 399–424. DOI: 10.1007/978-1-4939-1465-4_18 [51] Ward A. B., Sali A. and Wilson I. A. (2013). Integrative structural biology. Science 339:913−915. DOI:10.1126/science.1228565 [52] Yin R. and Pierce B. G. (2024). Evaluation of AlphaFold antibody–antigen modeling with implications for improving predictive accuracy. Protein Sci. 33:e4865. DOI:10.1002/pro.4865 [53] Noé F., Olsson S., Köhler J. et al. (2019). Boltzmann generators: Sampling equilibrium states of many-body systems with deep learning. Science 365:eaaw1147. DOI:10.1126/science.aaw1147 [54] Du Y. and Mordatch I. (2019). Implicit generation and modeling with energy based models. Adv. Neural Inf. Process. Syst. 32. [55] LeCun Y., Chopra S., Hadsell R. et al. (2006). A tutorial on energy-based learning. Predicting Structured Data 1. [56] Zhou H. and Zhou Y. (2002). Distance-scaled, finite ideal-gas reference state improves structure-derived potentials of mean force for structure selection and stability prediction. Protein Sci. 11:2714−2726. DOI:10.1110/ps.0217002 [57] Cunningham J. M., Koytiger G., Sorger P. K. et al. (2020). Biophysical prediction of protein–peptide interactions and signaling networks using machine learning. Nat. Methods 17:175−183. DOI:10.1038/s41592-019-0687-1 [58] Yan J. et al. (2024). Deltadock: a unified framework for accurate, efficient, and physically reliable molecular docking. Adv. Neural Inf. Process. Syst. 37:113282−113310. DOI:10.52202/079017-3599 [59] Li Y. et al. (2025). Decoding the limits of deep learning in molecular docking for drug discovery. Chem. Sci. 16:17374−17390. DOI:10.1039/d5sc05395a [60] Brown B. P. (2025). A generalizable deep learning framework for structure-based protein–ligand affinity ranking. Proc. Natl. Acad. Sci. U.S.A. 122:e2508998122. DOI:10.1073/pnas.2508998122 [61] Chang C.-e. A., Chen W. and Gilson M. K. (2007). Ligand configurational entropy and protein binding. Proc. Natl. Acad. Sci. U.S.A. 104:1534−1539. DOI:10.1073/pnas.0610494104 [62] Chodera J. D. and Mobley D. L. (2013). Entropy-enthalpy compensation: role and ramifications in biomolecular ligand recognition and design. Annu. Rev. Biophys. 42:121−142. DOI:10.1146/annurev-biophys-083012-130318 [63] Wu H. et al. (2024). in 12th Int. Conf. Learn. Represent. [64] Li J., Beaudoin C. and Ghosh S. (2023). Energy-based generative models for target-specific drug discovery. Front. Mol. Med. 3:1160877. DOI:10.3389/fmmed.2023.1160877 [65] Atz K., Grisoni F. and Schneider G. (2021). Geometric deep learning on molecular representations. Nat. Mach. Intell. 3:1023−1032. DOI:10.1038/s42256-021-00418-8 [66] Bronstein M. M. et al. (2021). Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. arXiv: 2104.13478. [67] Batzner S. et al. (2022). E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials. Nat. Commun. 13:2453. DOI:10.1038/s41467-022-29939-5 [68] Satorras V. G., Hoogeboom E. and Welling M. (2021). E(n) equivariant graph neural networks. In: Proc. 38th Int. Conf. Mach. Learn., pp. 9323–9332. [69] Schütt K. T., Unke O. T. and Gastegger M. (2021). Equivariant message passing for the prediction of tensorial properties and molecular spectra. In: Proc. 38th Int. Conf. Mach. Learn., pp. 9377–9388. [70] Stärk H., Ganea O.-E., Pattanaik L. et al. (2022). EquiBind: geometric deep learning for drug binding structure prediction. In: Proc. 39th Int. Conf. Mach. Learn., pp. 20503–20521. [71] Ganea O.-E. et al. (2021). Independent SE(3)-equivariant models for end-to-end rigid protein docking. arXiv: 2111.07786. [72] Hoogeboom E., Satorras V. G., Vignac C. et al. (2022). Equivariant diffusion for molecule generation in 3D. In: Proc. 39th Int. Conf. Mach. Learn., pp. 8867–8887. [73] Thomas N. et al. (2018). Tensor field networks: rotation- and translation-equivariant neural networks. arXiv: 1802.08219. [74] Fuchs F. B. et al. (2020). SE(3)-transformers: 3D roto-translation equivariant attention networks. Adv. Neural Inf. Process. Syst. 33:1970−1981. [75] Corso G., Stärk H., Jing B. et al. (2022). DiffDock: diffusion steps, twists, and turns for molecular docking. arXiv: 2210.01776. [76] Watson J. L. et al. (2023). De novo design of protein structure and function with RFdiffusion. Nature 620:1089−1100. DOI:10.1038/s41586-023-06415-8 [77] Batatia I., Kovacs D. P., Simm G. et al. (2022). MACE: higher order equivariant message passing neural networks. Adv. Neural Inf. Process. Syst. 35:11423−11436. DOI:10.52202/068431-0830 [78] Musaelian A. et al. (2023). Learning local equivariant representations for large-scale atomistic dynamics. Nat. Commun. 14:579. DOI:10.1038/s41467-023-36329-y [79] Haghighatlari M. et al. (2022). NewtonNet: a Newtonian message passing network for deep learning of interatomic potentials and forces. Digit. Discov. 1:333−343. DOI:10.1039/d2dd00008c [80] Schütt K. T. et al. (2017). SchNet: a continuous-filter convolutional neural network for modeling quantum interactions. Adv. Neural Inf. Process. Syst. 30. [81] Qiao Z., Nie W., Vahdat A. et al. (2024). State-specific protein–ligand complex structure prediction. Nat. Mach. Intell. 6:195−208. DOI:10.1038/s42256-024-00792-z [82] Lu W. et al. (2024). DynamicBind: predicting ligand-specific protein-ligand complex structure with a deep equivariant generative model. Nat. Commun. 15:1071. DOI:10.1038/s41467-024-45461-2 [83] Duval A. et al. (2023). A hitchhiker's guide to geometric GNNs for 3D atomic systems. arXiv: 2312.07511. [84] Geiger M. and Smidt T. (2022). e3nn: Euclidean neural networks. arXiv: 2207.09453. [85] Jing B., Eismann S., Suriana P. N. et al. (2020). Learning from protein structure with geometric vector perceptrons. arXiv: 2009.01411. [86] Ganea O.-E. et al. (2021). GeoMol: torsional geometric generation of molecular 3D conformer ensembles. Adv. Neural Inf. Process. Syst. 34:13757−13769. [87] Baek M. et al. (2021). Accurate prediction of protein structures and interactions using a three-track neural network. Science 373:871−876. DOI:10.1126/science.abj8754 [88] Krishna R. et al. (2024). Generalized biomolecular modeling and design with RoseTTAFold All-Atom. Science 384:eadl2528. DOI:10.1126/science.adl2528 DOI:10.1126/science.adl2528 [89] Baek M. et al. (2024). Accurate prediction of protein–nucleic acid complexes using RoseTTAFoldNA. Nat. Methods 21:117−121. DOI:10.1038/s41592-023-02086-5 [90] Jumper J. et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature 596:583−589. DOI:10.1038/s41586-021-03819-2 [91] Abramson J. et al. (2024). Accurate structure prediction of biomolecular interactions with AlphaFold 3. Nature 630:493−500. DOI:10.1038/s41586-024-07487-w [92] Chai Discovery. (2024). Chai-1: decoding the molecular interactions of life. bioRxiv 2024.10.10.615955. DOI:10.1101/2024.10.10.615955 [93] Ovchinnikov S. et al. (2017). Protein structure determination using metagenome sequence data. Science 355:294−298. DOI:10.1126/science.aah4043 [94] Hopf T. A. et al. (2014). Sequence co-evolution gives 3D contacts and structures of protein complexes. eLife 3:e03430. DOI:10.7554/eLife.03430 [95] Grisafi A., Wilkins D. M., Csányi G. et al. (2018). Symmetry-adapted machine learning for tensorial properties of atomistic systems. Phys. Rev. Lett. 120:036002. DOI:10.1103/PhysRevLett.120.036002 [96] Zhou G. et al. (2023). Uni-Mol: a universal 3D molecular representation learning framework. In: 11th Int. Conf. Learn. Represent.. [97] Pei Q. et al. (2023). FABind: fast and accurate protein-ligand binding. Adv. Neural Inf. Process. Syst. 36:55963−55980. DOI:10.52202/075280-2440 [98] AlQuraishi M. (2019). End-to-end differentiable learning of protein structure. Cell Syst. 8:292−301. DOI:10.1016/j.cels.2019.03.006 [99] Lee A. C.-L., Harris J. L., Khanna K. K. et al. (2019). A comprehensive review on current advances in peptide drug development and design. Int. J. Mol. Sci. 20:2383. DOI:10.3390/ijms20102383 [100] Fosgerau K. and Hoffmann T. (2015). Peptide therapeutics: current status and future directions. Drug Discov. Today 20:122−128. DOI:10.1016/j.drudis.2014.10.003 [101] Muttenthaler M. et al. (2021). Trends in peptide drug discovery. Nat. Rev. Drug Discov. 20:309−325. DOI:10.1038/s41573-020-00135-8 [102] Agoni C. et al. (2025). Molecular Modelling in Bioactive Peptide Discovery and Characterisation. Biomolecules 15:524. DOI:10.3390/biom15040524 [103] Wang J. et al. (2019). Improved modeling of peptide–protein binding through global docking and accelerated molecular dynamics simulations. Front. Mol. Biosci. 6:112. DOI:10.3389/fmolb.2019.00112 [104] Dutkiewicz Z. (2022). Computational methods for calculation of protein-ligand binding affinities in structure-based drug design. Phys. Sci. Rev. 7:933−968. DOI:10.1515/psr-2020-0034 [105] Seo S. et al. (2024). Pseq2Sites: enhancing protein sequence-based ligand binding-site prediction accuracy via the deep convolutional network and attention mechanism. Eng. Appl. Artif. Intell. 127:107257. DOI:10.1016/j.engappai.2023.107257 [106] Ciemny M. et al. (2018). Protein–peptide docking: opportunities and challenges. Drug Discov. Today 23:1530−1537. DOI:10.1016/j.drudis.2018.05.006 [107] Kilburg D. and Gallicchio E. (2016). Recent advances in computational models for the study of protein–peptide interactions. Adv. Protein Chem. Struct. Biol. 105:27−57. DOI:10.1016/bs.apcsb.2016.06.002 [108] Lee H. et al. (2015). GalaxyPepDock: a protein–peptide docking tool based on interaction similarity and energy optimization. Nucleic Acids Res. 43:W431−W435. DOI:10.1093/nar/gkv495 [109] Xue L. C. et al. (2015). Computational prediction of protein interfaces: A review of data driven methods. FEBS Lett. 589:3516−3526. DOI:10.1016/j.febslet.2015.10.003 [110] Rao R. et al. (2019). Evaluating protein transfer learning with TAPE. Adv. Neural Inf. Process. Syst. 32. [111] Zhao Z., Peng Z. and Yang J. (2018). Improving sequence-based prediction of protein–peptide binding residues by introducing intrinsic disorder and a consensus method. J. Chem. Inf. Model. 58:1459−1468. DOI:10.1021/acs.jcim.8b00019 [112] Vakser I. A. (2014). Protein-protein docking: From interaction to interactome. Biophys. J. 107:1785−1793. DOI:10.1016/j.bpj.2014.08.033 [113] Moal I. H., Moretti R., Baker D. et al. (2013). Scoring functions for protein–protein interactions. Curr. Opin. Struct. Biol. 23:862−867. DOI:10.1016/j.sbi.2013.06.017 [114] Hashemifar S., Neyshabur B., Khan A. A. et al. (2018). Predicting protein–protein interactions through sequence-based deep learning. Bioinformatics 34:i802−i810. DOI:10.1093/bioinformatics/bty573 [115] Yaffe M. B. (2002). Phosphotyrosine-binding domains in signal transduction. Nat. Rev. Mol. Cell Biol. 3:177−186. DOI:10.1038/nrm759 [116] Tinti M. et al. (2013). The SH2 domain interaction landscape. Cell Rep. 3:1293−1305. DOI:10.1016/j.celrep.2013.03.001 [117] Chen D. S. and Mellman I. (2013). Oncology meets immunology: the cancer-immunity cycle. Immunity 39:1−10. DOI:10.1016/j.immuni.2013.07.012 [118] Mellman I., Coukos G. and Dranoff G. (2011). Cancer immunotherapy comes of age. Nature 480:480−489. DOI:10.1038/nature10673 [119] Dunn G. P., Old L. J. and Schreiber R. D. (2004). The three Es of cancer immunoediting. Annu. Rev. Immunol. 22:329−360. DOI:10.1146/annurev.immunol.22.012703.104803 [120] Schumacher T. N. and Schreiber R. D. (2015). Neoantigens in cancer immunotherapy. Science 348:69−74. DOI:10.1126/science.aaa4971 [121] Gomase V. S., Sharma R. and Dhamane S. P. (2025). Innovative immunoinformatics tools for enhancing MHC class I epitope prediction. Protein Pept. Lett.. DOI:10.2174/0109298665373152250625054723 DOI:10.2174/0109298665373152250625054723 [122] Liu Y., Ouyang X.-H., Xiao Z.-X. et al. (2020). A review on the methods of peptide–MHC binding prediction. Curr. Bioinform. 15:878−888. DOI:10.2174/1574893615999200429122801 [123] Nielsen M. et al. (2007). NetMHCpan, a method for quantitative predictions of peptide binding to any HLA-A and -B locus protein. PLoS ONE 2:e796. DOI:10.1371/journal.pone.0000796 [124] Feltkamp M. C. W. et al. (1994). Efficient MHC class I-peptide binding is required but does not ensure MHC class I-restricted immunogenicity. Mol. Immunol. 31:1391−1401. DOI:10.1016/0161-5890(94)90155-4 [125] Murugan A. et al. (2012). Statistical inference of the generation probability of T-cell receptors from sequence repertoires. Proc. Natl. Acad. Sci. U.S.A. 109:16161−16166. DOI:10.1073/pnas.1212755109 [126] Dash P. et al. (2017). Quantifiable predictive features define epitope-specific T cell receptor repertoires. Nature 547:89−93. DOI:10.1038/nature22383 [127] Long X. et al. (2025). THLANet: a deep learning framework for predicting TCR–pHLA binding in immunotherapy applications. PLoS Comput. Biol. 21:e1013050. DOI:10.1371/journal.pcbi.1013050 [128] Atchley W. R. et al. (2005). Solving the protein sequence metric problem. Proc. Natl. Acad. Sci. U.S.A. 102:6395−6400. DOI:10.1073/pnas.0408677102 [129] Pettmann J. et al. (2021). The discriminatory power of the T cell receptor. eLife 10:e67092. DOI:10.7554/eLife.67092 [130] Cole D. K. et al. (2014). T-cell Receptor (TCR)-Peptide Specificity Overrides Affinity-enhancing TCR-Major Histocompatibility Complex Interactions. J. Biol. Chem. 289:628−638. DOI:10.1074/jbc.M113.522110 [131] Cuendet M. A., Zoete V. and Michielin O. (2011). How T cell receptors interact with peptide-MHCs: A multiple steered molecular dynamics study. Proteins 79:3007−3024. DOI:10.1002/prot.23104 [132] Liu Y. C. et al. (2013). Highly divergent T-cell receptor binding modes underlie specific recognition of a bulged viral peptide bound to a human leukocyte antigen class I molecule. J. Biol. Chem. 288:15442−15454. DOI:10.1074/jbc.M112.447185 [133] Albert B. A. et al. (2023). Deep Neural Networks Predict MHC-I Epitope Presentation and Transfer Learn Neoepitope Immunogenicity. bioRxiv. DOI:10.1101/2022.08.29.505690 DOI:10.1101/2022.08.29.505690 [134] Weber A., Born J. and Martínez M. R. (2021). TITAN: T-cell receptor specificity prediction with bimodal attention networks. Bioinformatics 37:i237−i244. DOI:10.1093/bioinformatics/btab294 [135] Hudson D., Fernandes R. A., Basham M. et al. (2023). Can we predict T cell specificity with digital biology and machine learning?. Nat. Rev. Immunol. 23:511−521. DOI:10.1038/s41577-023-00835-3 [136] Jurtz V. I. et al. (2018). NetTCR: sequence-based prediction of TCR binding to peptide-MHC complexes using convolutional neural networks. bioRxiv. DOI:10.1101/433706 DOI:10.1101/433706 [137] Altman J. D. et al. (1996). Phenotypic analysis of antigen-specific T lymphocytes. Science 274:94−96. DOI:10.1126/science.274.5284.94 [138] Zhang S.-Q. et al. (2018). High-throughput determination of the antigen specificities of T cell receptors in single cells. Nat. Biotechnol. 36:1156−1159. DOI:10.1038/nbt.4282 [139] Zhang Z. et al. (2021). Mapping the functional landscape of T cell receptor repertoires by single-T cell transcriptomics. Nat. Methods 18:92−99. DOI:10.1038/s41592-020-01020-3 [140] Tickotsky N. et al. (2017). McPAS-TCR: a manually curated catalogue of pathology-associated T cell receptor sequences. Bioinformatics 33:2924−2929. DOI:10.1093/bioinformatics/btx286 [141] Yamasaki S. et al. (2006). Mechanistic basis of pre-T cell receptor-mediated autonomous signaling critical for thymocyte development. Nat. Immunol. 7:67. DOI:10.1038/ni1290 [142] Reck M. et al. (2016). Pembrolizumab versus chemotherapy for PD-L1–positive non–small-cell lung cancer. N. Engl. J. Med. 375:1823−1833. DOI:10.1056/NEJMoa1606774 [143] Rizvi N. A. et al. (2015). Mutational landscape determines sensitivity to PD-1 blockade in non–small cell lung cancer. Science 348:124−128. DOI:10.1126/science.aaa1348 [144] Miao D. et al. (2018). Genomic correlates of response to immune checkpoint therapies in clear cell renal cell carcinoma. Science 359:801−806. DOI:10.1126/science.aan5951 [145] Yost K. E. et al. (2019). Clonal replacement of tumor-specific T cells following PD-1 blockade. Nat. Med. 25:1251−1259. DOI:10.1038/s41591-019-0522-3 [146] Lin Z. et al. (2023). Evolutionary-scale prediction of atomic-level protein structure with a language model. Science 379:1123−1130. DOI:10.1126/science.ade2574 [147] Elnaggar A. et al. (2023). Ankh: optimized protein language model unlocks general-purpose modelling. arXiv: 2301.06568. [148] Zhang S., Liu Y. and Xie L. (2022). Physics-aware graph neural network for accurate RNA 3D structure prediction. arXiv: 2210.16392. [149] Eraslan G., Avsec Ž., Gagneur J. et al. (2019). Deep learning: new computational modelling techniques for genomics. Nat. Rev. Genet. 20:389−403. DOI:10.1038/s41576-019-0122-6 [150] Boehr D. D., Nussinov R. and Wright P. E. (2009). The role of dynamic conformational ensembles in biomolecular recognition. Nat. Chem. Biol. 5:789−796. DOI:10.1038/nchembio.232 [151] Changeux J.-P. and Edelstein S. (2011). Conformational selection or induced fit? 50 years of debate resolved. F1000 Biol. Rep. 3:19. DOI:10.3410/B3-19 [152] Cocco S. et al. (2018). Inverse statistical physics of protein sequences: a key issues review. Rep. Prog. Phys. 81:032601. DOI:10.1088/1361-6633/aa9965 [153] Marks D. S., Hopf T. A. and Sander C. (2012). Protein structure prediction from sequence variation. Nat. Biotechnol. 30:1072−1080. DOI:10.1038/nbt.2419 [154] Duan G. and Walther D. (2015). The roles of post-translational modifications in the context of protein interaction networks. PLoS Comput. Biol. 11:e1004049. DOI:10.1371/journal.pcbi.1004049 [155] Wang J. et al. (2019). Machine learning of coarse-grained molecular dynamics force fields. ACS Cent. Sci. 5:755−767. DOI:10.1021/acscentsci.8b00913 [156] Gainza P. et al. (2020). Deciphering interaction fingerprints from protein molecular surfaces using geometric deep learning. Nat. Methods 17:184−192. DOI:10.1038/s41592-019-0666-6 [157] Preuer K. et al. (2019). Interpretable deep learning in drug discovery. In: Samek W., Montavon G., Vedaldi A. et al. (eds.), Explainable AI: Interpreting, Explaining and Visualizing Deep Learning (Springer), pp. 331–345. DOI: 10.1007/978-3-030-28954-6_18 [158] Mardt A., Pasquali L., Wu H. et al. (2018). VAMPnets for deep learning of molecular kinetics. Nat. Commun. 9:5. DOI:10.1038/s41467-017-02388-1 [159] Eden E. et al. (2009). GOrilla: a tool for discovery and visualization of enriched GO terms in ranked gene lists. BMC Bioinformatics 10:48. DOI:10.1186/1471-2105-10-48 [160] Noid W. G. (2013). Perspective: coarse-grained models for biomolecular systems. J. Chem. Phys. 139:090901. DOI:10.1063/1.4818908 [161] Zhang L. et al. (2018). Deep potential molecular dynamics: a scalable model with the accuracy of quantum mechanics. Phys. Rev. Lett. 120:143001. DOI:10.1103/PhysRevLett.120.143001 [162] Doerr S. et al. (2021). TorchMD: A deep learning framework for molecular simulations. J. Chem. Theory Comput. 17:2355−2363. DOI:10.1021/acs.jctc.0c01343 [163] Liu A. et al. (2024). DeepSeek-V3 Technical Report. arXiv: 2412.19437. [164] Gazizov A. et al. (2026). AF2BIND: predicting small-molecule binding sites using the pair representation of AlphaFold2. Nat. Methods 1−10. DOI:10.1038/s41592-026-03011-2 [165] Nielsen J. C., Hjorringgaard C., Nygaard M. M. et al. (2024). Machine-learning-guided peptide drug discovery: development of GLP-1 receptor agonists with improved drug properties. J. Med. Chem. 67:11814−11826. DOI:10.1021/acs.jmedchem.4c00417 -
ABOUT THIS ARTICLE
Cite this article:
Xie H., Wang H., Yao X., et al. (2026). Physics-informed deep learning: A new paradigm for macromolecular recognition and rational protein design. The Innovation Drug Discovery 1:100033. https://doi.org/10.59717/j.xinn-drugdisc.2026.100033
Welcome!
To request copyright permission to republish or share portions of our works, please visit Copyright Clearance Center's (CCC) Marketplace website at marketplace.copyright.com.
Export File
Citation
Xie H., Wang H., Yao X., et al. (2026). Physics-informed deep learning: A new paradigm for macromolecular recognition and rational protein design. The Innovation Drug Discovery 1:100033. https://doi.org/10.59717/j.xinn-drugdisc.2026.100033
Format
Content
-
Figure 1.
Conceptual prelude: The critical crossroads and paradigm shift in biomolecular modeling
-
Figure 2.
The global framework of Physics-Informed Deep Learning
-
Figure 3.
Core algorithmic mechanisms of the Physics-Informed Deep Learning paradigm
-
Figure 4.
Translational applications of the PIDL paradigm across diverse biological scales and therapeutic modalities
-
Figure 5.
Ontological restructuring: Overcoming systemic mismatches and mapping the future of biomolecular AI
