Searcharxiv⌕ Search

arXiv subjects

Soojung Yang

Publications and source records attributed to Soojung Yang.

14 recordsLinked to original sources

Science Prediction Universe

Humanity has long sought to explain natural phenomena. Reliable prediction of experimental outcomes is a central test of our understanding of the processes that produce them. If artificial general intelligence (AGI) is to surpass human scientific capabilities, it must also predict experimental outcomes more reliably than humans. Such capabilities could accelerate scientific discovery across disciplines, help address humanity's most pressing challenges, and improve human well-being. We introduce Science Prediction Universe (SPU), a benchmark of hidden experimental outcomes with 100 public verified tasks from chemistry, materials science, biology and medicine, and 16 private tasks from unpublished wet-lab experiments. Given an experimental setup and candidate conditions, an agent works with any resources it needs to rank the candidates by a measured outcome. We evaluate thirteen agent configurations spanning twelve frontier models across two harnesses (Codex and Claude Code). Current agents remain far from saturating wet-lab outcome prediction, with substantial performance disparities across models: on the public benchmark, configurations separate into two distinct tiers, with task-adjusted Spearman correlations ranging from 0.227 for the lowest-performing configuration to 0.478 for the best. Analysis of 1,155 scored public trajectories (227,103 tool calls) suggests that the bottleneck lies in scientific judgment rather than tool execution: agents rely on shared monotone priors, agreeing far more with one another than with experimental reality (0.52 vs. 0.33), and frequently compute trivial physical or chemical proxies that do not influence the final outcome.

cs.AI↗

Generative modeling of intrinsically disordered protein regions by reinforcing sparse autoencoder features

Intrinsically disordered protein regions (IDRs) play central roles in cellular processes such as transcriptional regulation, signal transduction, and subcellular localization, yet their functional design remains challenging. Structure-based design methods do not readily apply to IDRs, and existing protein language models are trained on full-length protein sequences, thus learning a prior that is biased towards folded domains. Here, we present IDiom, an autoregressive protein language model trained on IDiom-DB, a dataset of 54 million predicted IDRs curated from the AlphaFold Database. IDiom generates diverse sequences that recapitulate the composition, patterning, motifs, and predicted disorder of natural IDRs. To control function-associated sequence patterns, we also introduce reinforcement learning with sparse autoencoder features (RL-SAE), a post-training method that rewards the generation of sequences that activate specified feature sets. Across eight IDR design tasks, RL-SAE sequences activate, on average, 90% of 30 targeted features, compared to 24% for activation steering. We demonstrate that RL-SAE improves the predicted subcellular localization and transcriptional activity of generated IDRs compared to steering and supervised fine-tuning, and enables features associated with distinct biological functions to be combined within individual sequences. Thus, IDiom and RL-SAE enable interpretable and composable IDR design through explicit control of function-associated sequence features. More broadly, RL-SAE could extend to other protein design settings where interpretable features provide useful design targets. Code is available at https://github.com/rotskoff-group/idiom.

cs.LG↗

LLM sequential decision making under uncertainty in biochemical domains

Large language models (LLMs) are increasingly used to drive scientific discovery. Understanding how LLMs make decisions from new data and memory of the literature is vital before trusting them to design experiments under tight experimental budgets. However, their decision strategies are invisible in the current performance scores used to evaluate research agents. Here, we benchmark five frontier LLMs in a Bayesian Optimization setting against published statistical baselines on seven combinatorial datasets spanning protein engineering, reaction optimization, molecular design, peptide self-assembly, and catalysis. Performance is paired with direct measurements of model beliefs and actions, enabling highly resolved behavior analysis. A prompt ablation that progressively strips context separates memorization from chemical reasoning and from bare categorical optimization. Prior chemical knowledge helps in expectation, but with high variance and occasionally even harms performance. No configuration tested decisively beats a mean statistical baseline across domains. Belief-movement and Martingale diagnostics, corrected here for a measurement-noise bias that mislabels rational agents as irrational, show that models overreact to incoming data rather than entrenching on their priors in the contexts studied here. Interestingly, while LLM actions are exploitative, models sincerely intend to explore and consistently act on that intent. This failure is a competence gap arising from context-stickiness. Removing in-context history restores exploration, indicating that priors and data must be decoupled to achieve effective LLM-driven discovery.

cs.AI↗

VibeProteinBench: An Evaluation Benchmark for Language-interfaced Vibe Protein Design

Protein design aims to compose amino-acid sequences that fold into stable three-dimensional structures while satisfying targeted functional properties. The field is increasingly shifting toward vibe protein design, where a single model is expected to generate novel sequences, engineer existing proteins, and reason about protein characteristics through flexible natural-language constraints. However, existing benchmarks typically evaluate isolated aspects of protein design or assume predefined structured inputs, making them ill-suited to assess this broad, open-ended setting. To address this gap, we present Vibe Protein Design Benchmark (VibeProteinBench), a language-interfaced benchmark that evaluates whether a single model can operate across three complementary stages of a computational protein-design workflow: recognition, engineering, and generation. Each stage is grounded in expert-curated mechanistic rationales and multi-faceted in silico validation, to computationally verify whether model outputs are biologically plausible. Across diverse general-purpose and domain-specialized LLMs, evaluated with and without tool access, no model performs strongly across all three stages at once. Beyond measuring end-task success, VibeProteinBench exposes where current systems break: we identify recurring failure patterns and characterize how tool-augmented agents select tool arguments. Together, these findings show that vibe protein design remains a substantial open challenge.

q-bio.QM↗

A Priori Sampling of Transition States with Guided Diffusion

Transition states, the first-order saddle points on the potential energy surfaces, govern the kinetics and mechanisms of chemical reactions and conformational changes. Locating them is challenging because transition pathways are topologically complex and can proceed via an ensemble of diverse routes. Existing methods address these challenges by introducing heuristic assumptions about the pathway or reaction coordinates, which limits their applicability when a good initial guess is unavailable or when the guess precludes alternative, potentially relevant pathways. We propose to bypass such heuristic limitations by introducing ASTRA, A Priori Sampling of TRAnsition States with Guided Diffusion, which reframes the transition state search as an inference-time scaling problem for generative models. ASTRA trains a score-based diffusion model on configurations from known metastable states. Then, ASTRA guides inference toward the isodensity surface separating the basins of metastable states via a principled composition of conditional scores. A Score-Aligned Ascent (SAA) process then approximates a reaction coordinate from the difference between conditioned scores and combines it with physical forces to drive convergence onto first-order transition states. Validated on benchmarks ranging from 2D potentials to biomolecular conformational changes and a chemical reaction, ASTRA locates transition states with high precision and discovers multiple reaction pathways, enabling mechanistic studies of complex molecular systems.

physics.chem-ph↗

Learning Collective Variables from BioEmu with Time-Lagged Generation

Molecular dynamics is crucial for understanding molecular systems but its applicability is often limited by the vast timescales of rare events like protein folding. Enhanced sampling techniques overcome this by accelerating the simulation along key reaction pathways, which are defined by collective variables (CVs). However, identifying effective CVs that capture the slow, macroscopic dynamics of a system remains a major bottleneck. This work proposes a novel framework coined BioEmu-CV that learns these essential CVs automatically from BioEmu, a recently proposed foundation model for generating protein equilibrium samples. In particular, we re-purpose BioEmu to learn time-lagged generation conditioned on the learned CV, i.e., predict the distribution of molecular states after a certain amount of time. This training process promotes the CV to encode only the slow, long-term information while disregarding fast, random fluctuations. We validate our learned CV on fast-folding proteins with two key applications: (1) estimating free energy differences using on-the-fly probability enhanced sampling and (2) sampling transition paths with steered molecular dynamics. Our empirical study also serves as a new systematic and comprehensive benchmark for MLCVs on fast-folding proteins larger than Alanine Dipeptide.

cs.LG↗

Universally Converging Representations of Matter Across Scientific Foundation Models

Machine learning models of vastly different modalities and architectures are being trained to predict the behavior of molecules, materials, and proteins. However, it remains unclear whether they learn similar internal representations of matter. Understanding their latent structure is essential for building scientific foundation models that generalize reliably beyond their training domains. Although representational convergence has been observed in language and vision, its counterpart in the sciences has not been systematically explored. Here, we show that representations learned by nearly sixty scientific models, spanning string-, graph-, 3D atomistic, and protein-based modalities, are highly aligned across a wide range of chemical systems. Models trained on different datasets have highly similar representations of small molecules, and machine learning interatomic potentials converge in representation space as they improve in performance, suggesting that foundation models learn a common underlying representation of physical reality. We then show two distinct regimes of scientific models: on inputs similar to those seen during training, high-performing models align closely and weak models diverge into local sub-optima in representation space; on vastly different structures from those seen during training, nearly all models collapse onto a low-information representation, indicating that today's models remain limited by training data and inductive bias and do not yet encode truly universal structure. Our findings establish representational alignment as a quantitative benchmark for foundation-level generality in scientific models. More broadly, our work can track the emergence of universal representations of matter as models scale, and for selecting and distilling models whose learned representations transfer best across modalities, domains of matter, and scientific tasks.

cs.LG↗

Flow Matching for Accelerated Simulation of Atomic Transport in Crystalline Materials

Atomic transport underpins the performance of materials in technologies such as energy storage and electronics, yet its simulation remains computationally demanding. In particular, modeling ionic diffusion in solid-state electrolytes (SSEs) requires methods that can overcome the scale limitations of traditional ab initio molecular dynamics (AIMD). We introduce LiFlow, a generative framework to accelerate MD simulations for crystalline materials that formulates the task as conditional generation of atomic displacements. The model uses flow matching, with a Propagator submodel to generate atomic displacements and a Corrector to locally correct unphysical geometries, and incorporates an adaptive prior based on the Maxwell-Boltzmann distribution to account for chemical and thermal conditions. We benchmark LiFlow on a dataset comprising 25-ps trajectories of lithium diffusion across 4,186 SSE candidates at four temperatures. The model obtains a consistent Spearman rank correlation of 0.7-0.8 for lithium mean squared displacement (MSD) predictions on unseen compositions. Furthermore, LiFlow generalizes from short training trajectories to larger supercells and longer simulations while maintaining high accuracy. With speed-ups of up to 600,000$\times$ compared to first-principles methods, LiFlow enables scalable simulations at significantly larger length and time scales.

cond-mat.mtrl-sci↗

Transferable Learning of Reaction Pathways from Geometric Priors

Identifying minimum-energy paths (MEPs) is crucial for understanding chemical reaction mechanisms but remains computationally demanding. We introduce MEPIN, a scalable machine-learning method for efficiently predicting MEPs from reactant and product configurations, without relying on transition-state geometries or pre-optimized reaction paths during training. The task is defined as predicting deviations from geometric interpolations along reaction coordinates. We address this task with a continuous reaction path model based on a symmetry-broken equivariant neural network that generates a flexible number of intermediate structures. The model is trained using an energy-based objective, with efficiency enhanced by incorporating geometric priors from geodesic interpolation as initial interpolations or pre-training objectives. Our approach generalizes across diverse chemical reactions and achieves accurate alignment with reference intrinsic reaction coordinates, as demonstrated on various small molecule reactions and [3+2] cycloadditions. Our method enables the exploration of large chemical reaction spaces with efficient, data-driven predictions of reaction pathways.

physics.chem-ph↗

Learning Collective Variables with Synthetic Data Augmentation through Physics-Inspired Geodesic Interpolation

In molecular dynamics simulations, rare events, such as protein folding, are typically studied using enhanced sampling techniques, most of which are based on the definition of a collective variable (CV) along which acceleration occurs. Obtaining an expressive CV is crucial, but often hindered by the lack of information about the particular event, e.g., the transition from unfolded to folded conformation. We propose a simulation-free data augmentation strategy using physics-inspired metrics to generate geodesic interpolations resembling protein folding transitions, thereby improving sampling efficiency without true transition state samples. This new data can be used to improve the accuracy of classifier-based methods. Alternatively, a regression-based learning scheme for CV models can be adopted by leveraging the interpolation progress parameter.

physics.chem-ph↗

Chemically Transferable Generative Backmapping of Coarse-Grained Proteins

Coarse-graining (CG) accelerates molecular simulations of protein dynamics by simulating sets of atoms as singular beads. Backmapping is the opposite operation of bringing lost atomistic details back from the CG representation. While machine learning (ML) has produced accurate and efficient CG simulations of proteins, fast and reliable backmapping remains a challenge. Rule-based methods produce poor all-atom geometries, needing computationally costly refinement through additional simulations. Recently proposed ML approaches outperform traditional baselines but are not transferable between proteins and sometimes generate unphysical atom placements with steric clashes and implausible torsion angles. This work addresses both issues to build a fast, transferable, and reliable generative backmapping tool for CG protein representations. We achieve generalization and reliability through a combined set of innovations: representation based on internal coordinates; an equivariant encoder/prior; a custom loss function that helps ensure local structure, global structure, and physical constraints; and expert curation of high-quality out-of-equilibrium protein data for training. Our results pave the way for out-of-the-box backmapping of coarse-grained simulations for arbitrary proteins.

cs.LG↗

PIGNet: A physics-informed deep learning model toward generalized drug-target interaction predictions

Recently, deep neural network (DNN)-based drug-target interaction (DTI) models were highlighted for their high accuracy with affordable computational costs. Yet, the models' insufficient generalization remains a challenging problem in the practice of in-silico drug discovery. We propose two key strategies to enhance generalization in the DTI model. The first is to predict the atom-atom pairwise interactions via physics-informed equations parameterized with neural networks and provides the total binding affinity of a protein-ligand complex as their sum. We further improved the model generalization by augmenting a broader range of binding poses and ligands to training data. We validated our model, PIGNet, in the comparative assessment of scoring functions (CASF) 2016, demonstrating the outperforming docking and screening powers than previous methods. Our physics-informing strategy also enables the interpretation of predicted affinities by visualizing the contribution of ligand substructures, providing insights for further ligand optimization.

q-bio.BM↗

Hit and Lead Discovery with Explorative RL and Fragment-based Molecule Generation

Recently, utilizing reinforcement learning (RL) to generate molecules with desired properties has been highlighted as a promising strategy for drug design. A molecular docking program - a physical simulation that estimates protein-small molecule binding affinity - can be an ideal reward scoring function for RL, as it is a straightforward proxy of the therapeutic potential. Still, two imminent challenges exist for this task. First, the models often fail to generate chemically realistic and pharmacochemically acceptable molecules. Second, the docking score optimization is a difficult exploration problem that involves many local optima and less smooth surfaces with respect to molecular structure. To tackle these challenges, we propose a novel RL framework that generates pharmacochemically acceptable molecules with large docking scores. Our method - Fragment-based generative RL with Explorative Experience replay for Drug design (FREED) - constrains the generated molecules to a realistic and qualified chemical space and effectively explores the space to find drugs by coupling our fragment-based generation method and a novel error-prioritized experience replay (PER). We also show that our model performs well on both de novo and scaffold-based schemes. Our model produces molecules of higher quality compared to existing methods while achieving state-of-the-art performance on two of three targets in terms of the docking scores of the generated molecules. We further show with ablation studies that our method, predictive error-PER (FREED(PE)), significantly improves the model performance.

q-bio.QM↗

A comprehensive study on the prediction reliability of graph neural networks for virtual screening

Prediction models based on deep neural networks are increasingly gaining attention for fast and accurate virtual screening systems. For decision makings in virtual screening, researchers find it useful to interpret an output of classification system as probability, since such interpretation allows them to filter out more desirable compounds. However, probabilistic interpretation cannot be correct for models that hold over-parameterization problems or inappropriate regularizations, leading to unreliable prediction and decision making. In this regard, we concern the reliability of neural prediction models on molecular properties, especially when models are trained with sparse data points and imbalanced distributions. This work aims to propose guidelines for training reliable models, we thus provide methodological details and ablation studies on the following train principles. We investigate the effects of model architectures, regularization methods, and loss functions on the prediction performance and reliability of classification results. Moreover, we evaluate prediction reliability of models on virtual screening scenario. Our result highlights that correct choice of regularization and inference methods is evidently important to achieve high success rate, especially in data imbalanced situation. All experiments were performed under a single unified model implementation to alleviate external randomness in model training and to enable precise comparison of results.

cs.LG↗