Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· 0 citations· 15 references
TL;DR
Semantics-Preserving Exploratory Sampling (SPES), which integrates Graph Laplacian Spectral Noise to respect chain connectivity and Conditional Freedom Modulation to dynamically balance exploration with fidelity, enables diversity-enhanced generation without sacrificing geometric validity under the reported structural metrics, improving the empirical exploration--fidelity trade-off.
Abstract
Biomolecular binder design for peptides and antibodies requires generating diverse candidates that satisfy stringent three-dimensional geometric constraints while enabling affinity-oriented exploration under strong structural priors. In current generative models, the effective search space for structurally feasible binders is severely constrained, as the complexity of biochemical interactions is not explicitly encoded into a semantically grounded representation of viable molecular manifolds. To address this challenge, we propose Pretrained Representation Induced Molecular gEneration (PRIME), a unified generative framework for three-dimensional binder design across peptides and antibodies. PRIME grounds stochastic generation on frozen large-scale pretrained structural representations, inheriting robust physical priors to ensure structural feasibility without training a manifold from scratch. However, defining a feasible space alone is insufficient for effective exploration. Under commonly used isotropic perturbations, chain topology is ignored, allowing local noise to propagate into global structural distortions. To enable controlled exploration within the feasible space, we introduce Semantics-Preserving Exploratory Sampling (SPES), which integrates Graph Laplacian Spectral Noise to respect chain connectivity and Conditional Freedom Modulation to dynamically balance exploration with fidelity. By aligning stochastic exploration with structural semantics, PRIME enables diversity-enhanced generation without sacrificing geometric validity under the reported structural metrics, improving the empirical exploration--fidelity trade-off. PRIME achieves state-of-the-art performance on unified peptide and antibody benchmarks, effectively reconciling geometric validity with functional optimization under computational proxy metrics. The source code is available at https://github.com/simplaj/PRIME.
This work proposes a novel approach that learns hierarchical discrete representations of protein structures using vector quantization, and outperforms state-of-the-art models such as ESMDiff across challenging benchmark datasets, including BPTI MD trajectories and conformational-changing pairs.
Seokjun On, Yujin Jeong, Kanghyeon Kim et al.· Bioinformatics· 0 citations
Crystal Structure Prediction (CSP), the task of determining stable atomic arrangements from chemical composition alone, remains a central challenge in computational materials science with direct implications for accelerating materials discovery. While recent diffusion-based generative models achieve impressive results by conditioning on space-group symmetry information, this paradigm exposes three fundamental challenges: (1) unreliable symmetry inference from composition, (2) reliance on symmetry-only priors without comprehensive structural geometric guidance, and (3) prior-induced distribution shift caused by inaccurate or overly strong constraints. To address these challenges, we propose TGD-CSP, a three-stage generative framework that: (1) learns a cross-modal embedding space to retrieve structurally relevant templates directly from composition, thereby providing reliable symmetry priors; (2) guides diffusion-based generation via score-based conditioning that explicitly incorporates comprehensive geometric information from retrieved templates; and (3) fine-tunes the generative policy via reinforcement learning with an energy-based reward to alleviate prior-induced distribution shift and mitigate biased generation. % Evidence: Concrete results TGD-CSP achieves match rates of 74.56% and 75.25% on Perov-5 and MP-20, respectively, including a 42% relative improvement over state-of-the-art methods on Perov-5. It reduces RMSE to 0.0259 and 0.0258, respectively, while yielding structures with formation energies closely matching those of ground-truth structures. Our results demonstrate that TGD-CSP enables reliable generation from composition alone, significantly improving the practicality of generative models for real materials discovery.
Lu Yang, Tiantian Xu, X. Liu et al.· Proceedings of the 32nd ACM...· 0 citations
Molecular crystal structure prediction (CSP) is important in pharmaceuticals, agrochemicals, and organic electronics, where subtle differences in molecular conformation and packing can strongly affect material properties. We present Packora, a flow-based generative model for molecular CSP that jointly predicts atomic coordinates and the lattice from molecular graphs. Packora supports multi-component and organometallic crystals and can condition on any subset of molecular conformers, stereochemical labels, and space-group information within a single model. Inspired by the CCDC CSP blind test, we evaluate generation and ranking separately, using generation to isolate generator quality and ranking to measure end-to-end performance under a common relaxation and ranking pipeline. We also systematically study architecture, training, conditioning, inference, and scaling, identifying an effective design based on cacheable pairwise reasoning, training objective and numerical solver choices, conditioning dropout, and balanced scaling of pairwise and single representations. Packora outperforms the baselines on both structure generation and ranking benchmarks, achieving the best matched-budget coverage across all six generation benchmarks, as well as higher experimental-form recovery, lower experimental-form ranks, and faster convergence in ranking.
Nayoung Kim, Kiyoung Seong, Sungsoo Ahn· 0 citations
PINS (Physics-Informed NMR Structure elucidation model), a generative framework that explicitly bridges the gap between spectral data and molecular topology by enforcing multiphysical priors, provides a trustworthy, automated strategy for decoding novel chemical structures in data-scarce regimes.
Chemically modified and macrocyclic peptides are increasingly important therapeutics, yet current molecular representation models do not natively represent chemical modification and covalent topology in a unified way. Atom-level strings obscure macrocyclic connectivity, whereas protein sequence models cannot encode noncanonical residues and explicit cross-links. Here we pretrain an encoder-only transformer directly on Hierarchical Editing Language for Macromolecules (HELM) notation, which specifies monomer identity and connectivity. In this work, we show that the resulting representations achieve best mean performance in cyclic peptide membrane permeability prediction (random split R 2 = 0.668; retaining best mean performance under a Murcko scaffold split), exceeding external pretrained SMILES-based encoders. An architecture-matched SMILES control narrowed the HELM–SMILES gap under full finetuning, whereas HELM-BERT retained clearer advantages in frozen-representation settings. HELM-BERT also preserves HELM-specified macrocyclic topology in a linearly accessible form and supports competitive peptide–protein interaction prediction across complementary Propedia and ChEMBL benchmarks. More broadly, these findings suggest that pretraining directly on notation that makes structural constraints explicit offers a transferable strategy for biomolecular modalities that fall between small-molecule chemistry and protein sequence.
Seungeon Lee, Takuto Koyama, Itsuki Maeda et al.· Journal of Chemical Informat...· 0 citations