NeurIPS 2025 Papers

Virtual screening (VS) is a critical component of modern drug discovery, yet most existing methods—whether physics-based or deep learning-based—are developed around *holo* protein structures with known ligand-bound pockets. Consequently, their performance degrades significantly on *apo* or predicted structures such as those from AlphaFold2, which are more representative of real-world early-stage drug discovery, where pocket information is often missing. In this paper, we introduce an alignment-and-aggregation framework to enable accurate virtual screening under structural uncertainty. Our method comprises two core components: (1) a tri-modal contrastive learning module that aligns representations of the ligand, the *holo* pocket, and cavities detected from structures, thereby enhancing robustness to pocket localization error; and (2) a cross-attention based adapter for dynamically aggregating candidate binding sites, enabling the model to learn from activity data even without precise pocket annotations. We evaluated our method on a newly curated benchmark of *apo* structures, where it significantly outperforms state-of-the-art methods in blind apo setting, improving the early enrichment factor (EF1\%) from 11.75 to 37.19. Notably, it also maintains strong performance on *holo* structures. These results demonstrate the promise of our approach in advancing first-in-class drug discovery, particularly in scenarios lacking experimentally resolved protein-ligand complexes. Our implementation is publicly available at [https://github.com/Wiley-Z/AANet](https://github.com/Wiley-Z/AANet).

When language model (LM) users aim to improve the quality of its generations, it is crucial to specify concrete behavioral attributes that the model should strive to reflect. However, curating such principles across many domains, even non-exhaustively, requires a labor-intensive annotation process. To automate this process, we propose eliciting these latent attributes that guide model reasoning toward human-preferred responses by explicitly modeling them in a self-correction setting. Our approach mines new principles from the LM itself and compresses the discovered elements to an interpretable set via clustering. Specifically, we employ a form of posterior-regularized Monte Carlo Expectation-Maximization to both identify a condensed set of the most effective latent principles and teach the LM to strategically invoke them in order to intrinsically refine its responses. We demonstrate that bootstrapping our algorithm over multiple iterations enables smaller language models (7-8B parameters) to self-improve, achieving +8-10\% in AlpacaEval win-rate, an average of +0.3 on MT-Bench, and +19-23\% in principle-following win-rate on IFEval. We also show that clustering the principles yields interpretable and diverse model-generated constitutions while retaining model performance. The gains that our method achieves highlight the potential of automated, principle-driven post-training recipes toward continual self-improvement.

Predicting the bioactivity of candidate ligands remains a central challenge in drug discovery. Ligands and endogenous substrates often compete for the same binding sites on target proteins, and the extent to which a ligand can modulate protein function depends not only on its binding but also on how effectively it occupies the relevant pocket. However, most existing methods focus narrowly on local interactions within protein–ligand complexes and neglect spatial emptiness—the unoccupied regions within the binding site that may permit endogenous molecules to engage or interfere. Such unfilled space can diminish the ligand’s functional impact, regardless of binding affinity. To overcome this key limitation in protein–ligand modeling, we propose LigoSpace, a novel method integrating three core components. LigoSpace introduces GeoREC (Geometric Representation of Spatial Emptiness in Complexes) to quantify atomic-level empty space and Union-Pocket to unify multiple protein pockets, providing a global view of binding sites. Additionally, LigoSpace employs a pairwise loss instead of commonly used MSE loss, to better capture relative relationships critical for drug discovery. Extensive experiments on multiple datasets with diverse bioactivity types demonstrate that LigoSpace significantly improves performance when integrated into state-of-the-art models, highlighting the effectiveness of its novel components.

Predicting interactions between biochemical entities is a core challenge in drug discovery and systems biology, often hindered by limited data and poor generalization to unseen entities. Traditional discriminative models frequently underperform in such settings. We propose BioCG (Biochemical Constrained Generation), a novel framework that reformulates interaction prediction as a constrained sequence generation task. BioCG encodes target entities as unique discrete sequences via Iterative Residual Vector Quantization (I-RVQ) and trains a generative model to produce the sequence of an interacting partner given a query entity. A trie-guided constrained decoding mechanism, built from a catalog of valid target sequences, concentrates the model's learning on the critical distinctions between valid biochemical options, ensuring all outputs correspond to an entity within the pre-defined target catalog. An information-weighted training objective further focuses learning on the most critical decision points. BioCG achieves state-of-the-art (SOTA) performance across diverse tasks, Drug-Target Interaction (DTI), Drug-Drug Interaction (DDI), and Enzyme-Reaction Prediction, especially in data-scarce and cold-start conditions. On the BioSNAP DTI benchmark, for example, BioCG attains an AUC of 89.31\% on unseen proteins, representing a 14.3 percentage point gain over prior SOTA. By directly generating interacting partners from a known biochemical space, BioCG provides a robust and data-efficient solution for in-silico biochemical discovery.

Federated Generalized Category Discovery (Fed-GCD) requires a global model to classify seen classes and discover novel classes when data are siloed across heterogeneous clients. Existing GCD work often makes unrealistic assumptions, such as the need for prior knowledge of the number of novel classes or the assumption of uniform class distribution. We present Federated Local Prior Alignment (FedLPA), which eliminates these unrealistic assumptions by grounding learning in client-local structure and aligning predictions to client-local priors. Each client builds a similarity graph refined with reliable seen-class signals and discovers client-specific concepts and prototypes via Infomap. Leveraging the discovered concept structures, we introduce Local Prior Alignment (LPA): a self-distillation loss that matches the batch-mean prediction to an empirical prior computed from current concept assignments. The iterative process of local structure discovery and dynamic prior adaptation enables robust generalized category discovery under severe data heterogeneity. Our framework significantly outperforms existing federated generalized category discovery approaches on fine-grained and standard benchmarks, as demonstrated by extensive experimental results.

Large language models (LLMs) have shown promise in automating scientific hypothesis generation, yet existing approaches primarily yield coarse-grained hypotheses lacking critical methodological and experimental details. We introduce and formally define the new task of fine-grained scientific hypothesis discovery, which entails generating detailed, experimentally actionable hypotheses from coarse initial research directions. We frame this as a combinatorial optimization problem and investigate the upper limits of LLMs' capacity to solve it when maximally leveraged. Specifically, we explore four foundational questions: (1) how to best harness an LLM's internal heuristics to formulate the fine-grained hypothesis it itself would judge as the most promising among all the possible hypotheses it might generate, based on its own internal scoring-thus defining a latent reward landscape over the hypothesis space; (2) whether such LLM-judged better hypotheses exhibit stronger alignment with ground-truth hypotheses; (3) whether shaping the reward landscape using an ensemble of diverse LLMs of similar capacity yields better outcomes than defining it with repeated instances of the strongest LLM among them; and (4) whether an ensemble of identical LLMs provides a more reliable reward landscape than a single LLM. To address these questions, we propose a hierarchical search method that incrementally proposes and integrates details into the hypothesis, progressing from general concepts to specific experimental configurations. We show that this hierarchical process smooths the reward landscape and enables more effective optimization. Empirical evaluations on a new benchmark of expert-annotated fine-grained hypotheses from recent literature show that our method consistently outperforms strong baselines.

Foundation models (FMs), such as GPT-4 and AlphaFold, are reshaping the landscape of scientific research. Beyond accelerating tasks such as hypothesis generation, experimental design, and result interpretation, they prompt a more fundamental question: Are FMs merely enhancing existing scientific methodologies, or are they redefining the way science is conducted? In this paper, we argue that FMs are catalyzing a transition toward a new scientific paradigm. We introduce a three-stage framework to describe this evolution: (1) Meta-Scientific Integration, where FMs enhance workflows within traditional paradigms; (2) Hybrid Human-AI Co-Creation, where FMs become active collaborators in problem formulation, reasoning, and discovery; and (3) Autonomous Scientific Discovery, where FMs operate as independent agents capable of generating new scientific knowledge with minimal human intervention. Through this lens, we review current applications and emerging capabilities of FMs across existing scientific paradigms. We further identify risks and future directions for FM-enabled scientific discovery. This position paper aims to support the scientific community in understanding the transformative role of FMs and to foster reflection on the future of scientific discovery.

Generative modeling is increasingly important for data-driven computational design. Conventional approaches pair a generative model with a discriminative model to select or guide samples toward optimized designs. Yet discriminative models often struggle in data-scarce settings, common in scientific applications, and are unreliable in the tails of the distribution where optimal designs typically lie. We introduce generative property enhancer (GPE), an approach that implicitly guides generation by matching samples with lower property values to higher-value ones. Formulated as conditional density estimation, our framework defines a target distribution with improved properties, compelling the generative model to produce enhanced, diverse designs without auxiliary predictors. GPE is simple, scalable, end-to-end, modality-agnostic, and integrates seamlessly with diverse generative model architectures and losses. We demonstrate competitive empirical results on standard _in silico_ offline (non-sequential) protein fitness optimization benchmarks. Finally, we propose iterative training on a combination of limited real data and self-generated synthetic data, enabling extrapolation beyond the original property ranges.

The inverse folding problem, aiming to design amino acid sequences that fold into desired three-dimensional structures, is pivotal for various biotechnological applications. Here, we introduce a novel approach leveraging Direct Preference Optimization (DPO) to fine-tune an inverse folding model using feedback from a protein folding model. Given a target protein structure, we begin by sampling candidate sequences from the inverse‐folding model, then predict the three‐dimensional structure of each sequence with the folding model to generate pairwise structural‐preference labels. These labels are used to fine‐tune the inverse‐folding model under the DPO objective. Our results on the CATH 4.2 test set demonstrate that DPO fine-tuning not only improves sequence recovery of baseline models but also leads to a significant improvement in average TM-Score from 0.77 to 0.81, indicating enhanced structure similarity. Furthermore, iterative application of our DPO-based method on challenging protein structures yields substantial gains, with an average TM-Score increase of 79.5\% with regard to the baseline model. This work establishes a promising direction for enhancing protein sequence design ability from structure feedback by effectively utilizing preference optimization.

Large language model-assisted Evolutionary Search (LES) has emerged as a promising approach for Automated Algorithm Discovery (AAD). While many evolutionary search strategies have been developed for classic optimization problems, LES operates in abstract language spaces, presenting unique challenges for applying these strategies effectively. To address this, we propose a general LES framework that incorporates feature-assisted niche construction within abstract search spaces, enabling the seamless integration of niche-based search strategies from evolutionary computation. Building on this framework, we introduce PartEvo, an LES method that combines niche collaborative search and advanced prompting strategies to improve algorithm discovery efficiency. Experiments on both synthetic and real-world optimization problems show that PartEvo outperforms human-designed baselines and surpasses prior LES methods, such as Eoh and Funsearch. In particular, on resource scheduling tasks, PartEvo generates meta-heuristics with low design costs, achieving up to 90.1\% performance improvement over widely-used baseline algorithms, highlighting its potential for real-world applications.

Generalized Category Discovery (GCD) focuses on classifying known categories while simultaneously discovering novel categories from unlabeled data. However, previous GCD methods face challenges due to inconsistent optimization objectives and category confusion. This leads to feature overlap and ultimately hinders performance on novel categories. To address these issues, we propose the Neural Collapse-inspired Generalized Category Discovery (NC-GCD) framework. By pre-assigning and fixing Equiangular Tight Frame (ETF) prototypes, our method ensures an optimal geometric structure and a consistent optimization objective for both known and novel categories. We introduce a Consistent ETF Alignment Loss that unifies supervised and unsupervised ETF alignment and enhances category separability. Additionally, a Semantic Consistency Matcher (SCM) is designed to maintain stable and consistent label assignments across clustering iterations. Our method significantly enhancing novel category accuracy and demonstrating its effectiveness.

Inspired by the human ability to learn and organize knowledge into hierarchical taxonomies with prototypes, this paper addresses key limitations in current deep hierarchical clustering methods. Existing methods often tie the structure to the number of classes and underutilize the rich prototype information available at intermediate hierarchical levels. We introduce deep taxonomic networks, a novel deep latent variable approach designed to bridge these gaps. Our method optimizes a large latent taxonomic hierarchy, specifically a complete binary tree structured mixture-of-Gaussian prior within a variational inference framework, to automatically discover taxonomic structures and associated prototype clusters directly from unlabeled data without assuming true label sizes. We analytically show that optimizing the ELBO of our method encourages the discovery of hierarchical relationships among prototypes. Empirically, our learned models demonstrate strong hierarchical clustering performance, outperforming baselines across diverse image classification datasets using our novel evaluation mechanism that leverages prototype clusters discovered at all hierarchical levels. Qualitative results further reveal that deep taxonomic networks discover rich and interpretable hierarchical taxonomies, capturing both coarse-grained semantic categories and fine-grained visual distinctions.

Protein design is a fundamental challenge in biotechnology, aiming to design novel sequences with specific functions within the vast space of possible proteins. Recent advances in deep generative models have enabled function-based protein design from textual descriptions, yet struggle with structural plausibility. Inspired by classical protein design methods that leverage natural protein structures, we explore whether incorporating fragments from natural proteins can enhance foldability in generative models. Our empirical results show that even random incorporation of fragments improves foldability. Building on this insight, we introduce ProDVa, a novel protein design approach that integrates a text encoder for functional descriptions, a protein language model for designing proteins, and a fragment encoder to dynamically retrieve protein fragments based on textual functional descriptions. Experimental results demonstrate that our approach effectively designs protein sequences that are both functionally aligned and structurally plausible. Compared to state-of-the-art models, ProDVa achieves comparable function alignment using less than 0.04% of the training data, while designing significantly more well-folded proteins, with the proportion of proteins having pLDDT above 70 increasing by 7.38% and those with PAE below 10 increasing by 9.62%.

In multi-objective black-box optimization, the goal is typically to find solutions that optimize a set of $T$ black-box objective functions, $f_1, \ldots f_T$, simultaneously. Traditional approaches often seek a single Pareto-optimal set that balances trade-offs among all objectives. In contrast, we consider a problem setting that departs from this paradigm: finding a small set of $K < T$ solutions, that collectively "cover" the $T$ objectives. A set of solutions is defined as "covering" if, for each objective $f_1, \ldots f_T$, there is at least one good solution. A motivating example for this problem setting occurs in drug design. For example, we may have $T$ pathogens and aim to identify a set of $K < T$ antibiotics such that at least one antibiotic can be used to treat each pathogen. This problem, known as coverage optimization, has yet to be tackled with the Bayesian optimization (BO) framework. To fill this void, we develop Multi-Objective Coverage Bayesian Optimization (MOCOBO), a BO algorithm for solving coverage optimization. Our approach is based on a new acquisition function reminiscent of expected improvement in the vanilla BO setup. We demonstrate the performance of our method on high-dimensional black-box optimization tasks, including applications in peptide and molecular design. Results show that the coverage of the $K < T$ solutions found by MOCOBO matches or nearly matches the coverage of $T$ solutions obtained by optimizing each objective individually. Furthermore, in *in vitro* experiments, the peptides found by MOCOBO exhibited high potency against drug-resistant pathogens, further demonstrating the potential of MOCOBO for drug discovery. All of our code is publicly available at the following link: https://github.com/nataliemaus/mocobo.

Inferring synaptic connectivity from neural population activity is a fundamental challenge in computational neuroscience, complicated by partial observability and mismatches between inference models and true circuit dynamics. In this study, we propose a graph-based neural inference model that simultaneously predicts neural activity and infers latent connectivity by modeling neurons as interacting nodes in a graph. The architecture features two distinct modules: one for learning structural connectivity and another for predicting future spiking activity via a graph neural network (GNN). Our model accommodates unobserved neurons through auxiliary nodes, allowing for inference in partially observed circuits. We evaluate this approach using synthetic data generated from ring attractor network models and real spike recordings from head direction cells in mice. Across a wide range of conditions, including varying recurrent connectivity, external inputs, and incomplete observations, our model reliably resolves spurious correlations and recovers accurate weight profiles. When applied to real data, the inferred connectivity aligns with theoretical predictions of continuous attractor models. These results highlight the potential of GNN-based models to infer latent neural circuitry through self-supervised structure learning, while leveraging the spike prediction task to flexibly link connectivity and dynamics across both simulated and biological neural systems.

Oligonucleotide therapeutics offer great potential to address previously undruggable targets and enable personalized medicine. However, their progress is often hindered by insufficient safety and efficacy profiles. Predictive modeling and machine learning could significantly accelerate oligonucleotide drug discovery by identifying suboptimal compounds early on, but their application in this area lags behind other modalities. A key obstacle to the adoption of machine learning in the field is the scarcity of readily accessible and standardized datasets for model development, as data are often scattered across diverse experiments with inconsistent molecular representations. To overcome this challenge, we introduce OligoGym, a curated collection of standardized, machine learning-ready datasets encompassing various oligonucleotide therapeutic modalities and endpoints. We used OligoGym to benchmark diverse classical and deep learning methods, establishing performance baselines for each dataset across different featurization techniques, model configurations, and splitting strategies. Our work represents a crucial first step in creating a more unified framework for oligonucleotide therapeutic dataset generation and model training.

Scientific discovery across diverse fields increasingly grapples with datasets exhibiting pathological long-tailed distributions: a few common phenomena overshadow a multitude of rare yet scientifically critical instances. Unlike standard benchmarks, these scientific datasets often feature extreme imbalance coupled with a modest number of classes and limited overall sample volume, rendering existing long-tailed recognition (LTR) techniques ineffective. Such methods, biased by majority classes or prone to overfitting on scarce tail data, frequently fail to identify the very instances—novel materials, rare disease biomarkers, faint astronomical signals—that drive scientific breakthroughs. This paper introduces a novel, end-to-end framework explicitly designed to address pathological long-tailed recognition in scientific contexts. Our approach synergizes a Balanced Supervised Contrastive Learning (B-SCL) mechanism, which enhances the representation of tail classes by dynamically re-weighting their contributions, with a Smooth Objective Regularization (SOR) strategy that manages the inherent tension between tail-class focus and overall classification performance. We introduce and analyze the real-world ZincFluor chemical dataset ($\mathcal{T}=137.54$) and synthetic benchmarks with controllable extreme imbalances (CIFAR-LT variants). Extensive evaluations demonstrate our method's superior ability to decipher these extremes. Notably, on ZincFluor, our approach achieves a Tail Top-2 accuracy of $66.84\%$, significantly outperforming existing techniques. On CIFAR-10-LT with an imbalance ratio of $1000$ ($\mathcal{T}=100$), our method achieves a tail-class accuracy of $38.99\%$, substantially leading the next best. These results underscore our framework's potential to unlock novel insights from complex, imbalanced scientific datasets, thereby accelerating discovery.

We explore if it is possible to learn data structures end-to-end with neural networks, with a focus on the problem of nearest-neighbor (NN) search. We introduce a framework for data structure discovery, which adapts to the underlying data distribution and provides fine-grained control over query and space complexity. Crucially, the data structure is learned from scratch, and does not require careful initialization or seeding with candidate data structures. In several settings, we are able to reverse-engineer the learned data structures and query algorithms. For 1D nearest neighbor search, the model discovers optimal distribution (in)dependent algorithms such as binary search and variants of interpolation search. In higher dimensions, the model learns solutions that resemble k-d trees in some regimes, while in others, elements of locality-sensitive hashing emerge. Additionally, the model learns useful representations of high-dimensional data such as images and exploits them to design effective data structures. Beyond NN search, we believe the framework could be a powerful tool for data structure discovery for other problems and adapt our framework to the problem of estimating frequencies over a data stream. To encourage future work in this direction, we conclude with a discussion on some of the opportunities and remaining challenges of learning data structures end-to-end.

Predicting changes in binding free energy ($\Delta\Delta G$) is essential for understanding protein-protein interactions, which are critical in drug design and protein engineering. However, existing methods often rely on pre-trained knowledge and heuristic features, limiting their ability to accurately model complex mutation effects, particularly higher-order and many-body interactions. To address these challenges, we propose H3-DDG, a Hypergraph-driven Hierarchical network to capture Higher-order many-body interactions across multiple scales. By introducing a hierarchical communication mechanism, H3-DDG effectively models both local and global mutational effects. Experimental results demonstrate state-of-the-art performance on multiple benchmarks. On the SKEMPI v2 dataset, H3-DDG achieves a Pearson correlation of 0.75, improving multi-point mutations prediction by 12.10%. On the challenging BindingGYM dataset, it outperforms Prompt-DDG and BA-DDG by 62.61% and 34.26%, respectively. Ablation and efficiency analyses demonstrate its robustness and scalability, while a case study on SARS-CoV-2 antibodies highlights its practical value in improving binding affinity for therapeutic design.