Humans rapidly learn abstract knowledge when encountering novel environments and flexibly deploy this knowledge to guide efficient and intelligent action. Can modern AI systems learn and plan in a similar way? We study this question using a dataset of complex human gameplay with concurrent fMRI recordings, in which participants learn novel video games that require rule discovery, hypothesis revision, and multi-step planning. We jointly evaluate models by their ability to play the games, match human learning behavior, and predict brain activity during the same task, comparing a suite of frontier Large Reasoning Models (LRMs) against model-free and model-based deep reinforcement learning agents and a Bayesian theory-based agent. We find that frontier LRMs most closely match human behavioral patterns during game discovery and predict brain activity an order of magnitude better than both reinforcement learning alternatives across cortical and subcortical regions, with effects robust to permutation controls. Through targeted manipulations, we further show that brain alignment reflects the model's in-context representation of the game state rather than its downstream planning or reasoning. Our results establish LRMs as compelling computational accounts of human learning and decision making in complex, naturalistic environments. Project page with interactive replays: https://botcs.github.io/reason-to-play/

Oscillatory activity in auditory cortex is thought to play a central role in auditory and speech processing by synchronizing neural rhythms to external acoustic features of the speech stream. To support this function, cortical oscillators must flexibly phase-lock to inputs spanning a wide range of timescales, including rhythms substantially slower than their intrinsic frequency. Here we identify a general dynamical mechanism by which intrinsic inhibitory currents operating on multiple timescales enable such flexible phase-locking. Using tools from dynamical systems theory, we show that interactions between slow and superslow inhibitory processes generate prolonged post-input recovery delays through delayed Hopf phenomena, thereby substantially expanding the frequency range over which entrainment can occur. We demonstrate this mechanisms in a biophysically grounded cortical theta oscillator model for speech segmentation. Specifically, we show that both a theta-timescale (4-8 Hz) inhibitory current $I_m$ and a slower delta-timescale (1-4 Hz) inhibitory potassium current $I_{\rm K_{SS}}$ are crucial for entrainment flexibility. Their interaction creates a three-timescale structure that gives rise to pronounced delay phenomena associated with a delayed Hopf bifurcation (DHB). Interestingly, the superslow $I_{\rm K_{SS}}$ and the associated DHB play little role in the unforced oscillatory dynamics, but are recruited to support phase locking under external forcing. Moreover, the intermediate-timescale current $I_m$, rather than being redundant, further expands the phase-locking range by prolonging delayed recovery along the superslow manifold. Together, these results suggest that coordination among intrinsic inhibitory currents operating on multiple timescales may represent a key mechanism supporting flexible phase locking to rhythmic inputs in the brain.

Understanding how molecular alterations propagate across biological systems to drive disease remains a central challenge. Although high-throughput profiling enables comprehensive characterization of tumor states, most models neglect structured biological relationships or lack interpretability across scales. Here we present PPI-Net, a hierarchical graph neural network that integrates protein-protein interaction (PPI) networks with pathway-level representations to model disease from molecular interactions to functional processes. Patient-specific molecular profiles are embedded within a shared interaction network from STRING and propagated through a multi-layer Reactome hierarchy using graph attention, enabling aggregation of gene-level signals into higher-order biological programs. Across RNA-seq data from ten cancer types from The Cancer Genome Atlas, PPI-Net achieves robust predictive performance, with balanced accuracy exceeding 90% in multiple cohorts. Comparative analysis on RNA-Seq data from breast cancer demonstrated that PPI-Net's integration of the Reactome hierarchy improved balanced accuracy by 6.7% relative to a PPI-only model, while hierarchical multi-level supervision improved balanced accuracy by 12.3% relative to using only a single top-level prediction head. Applying a multi-omics approach using RNA-seq and methylation data improves model interpretation, recovering canonical oncogenic modules, including TP53-AKT signaling and stress response pathways, while revealing convergence onto coherent programs such as ion signaling and cellular responses to stimuli. These results demonstrate that integrating interaction networks with pathway hierarchies enables accurate prediction while providing mechanistic insight into cancer biology.

High-dimensional count data arise in applications such as single-cell RNA sequencing and neural spike trains, where mapping between distributions across successive batches or time points form critical components of data analysis. The recent success of diffusion- and flow-based deep generative models for images, video, and text motivates extending these ideas to count-valued settings, but many existing methods either treat each count as a categorical state or transform counts into a continuous space, neither of which is natural or efficient when the count range is large. We propose count-FM, a flow-matching framework for count data based on a continuous-time birth-death process with local unit jumps. Count-FM learns marginal transitions efficiently in count space through simulation-free training of conditional transition rates, allowing transport between arbitrary count-distributed source and target populations. In simulation, count-FM achieves better sample quality than representative baselines while using substantially fewer parameters. We further apply count-FM to scRNA-seq and neural spike-train data for unconditional generation, transport, and conditional generation. Across these tasks, count-FM yields improved sample quality, greater modeling efficiency, and interpretable transport paths.

This paper develops a predictive switching control algorithm for stochastic gene regulatory networks described by a Partial Integro-Differential Equation (PIDE) model, which enables direct shape control of the probability density function. Control inputs are selected from a finite candidate set to minimize a prescribed cost functional. A hybrid framework is proposed for scalability in higher-dimensional systems, using neural networks to approximate the control policy. A central theoretical contribution is a contraction-based analysis of the closed-loop PIDE dynamics. The paper establishes $L^ 1$-contractivity under the proposed control scheme, yielding formal stability guarantees and showing that the evolution of the probability density becomes progressively independent of the initial condition. Moreover, under strictly positive leakage terms, exponential convergence is obtained. The effectiveness and flexibility of the approach, together with the theoretical contractivity results, are illustrated through numerical simulations on three representative examples of increasing dimensionality.

RNA inverse sequence design has broad biological and engineering applications, but computational methods for practical design queries remain limited. Such queries may impose several constraints at once, including target folds or motifs, fixed bases, and coding restrictions, while leaving arbitrary sequence and structure in unspecified regions. Because these constraints may permit many acceptable sequences, we study RNA design as a conditional generative modeling problem. The basic object is a conditional law over RNA sequences given a user-specified condition, with full inverse folding as a special case. We introduce GoForth, a forward-trained RNA language model that conditions on structure, sequence, and coding targets. The formulation separates three ingredients that are often entangled in RNA design: a sequence prior, a forward folding sampler, and a reward or likelihood oracle. We train encoder-decoder models on witnessed folds rather than on outputs from an inverse-design teacher and validate our methodology on full inverse-folding benchmarks, as well as tasks involving constraints on structure, sequence, and coding. The resulting models achieve fast and high-quality candidate generation for mixed RNA design specifications. Moreover they furnish useful semantic embeddings of design tasks and a robust learned notion of designability.

Protein language models are trained primarily with masked language modeling (MLM), which predicts amino-acid identities at masked positions. We ask whether latent-space prediction can complement these token-level objectives under matched wall-clock budget. Across pretrained and random-init protein sequence encoders at 35--150M parameters, we find that the best protein-JEPA design is not all-position latent prediction but a variant: predicting latent targets only at masked positions, and retaining the MLM cross-entropy. We call this recipe masked-position MLM+JEPA. On a 16-task downstream suite (15 frozen linear probes plus SCOPe-40 zero-shot fold retrieval), under matched wall-clock budgets, this recipe wins more tasks than it loses against MLM-only continuation: 10 wins / 3 losses / 3 ties (hereafter W/L/T) on pretrained ESM2-35M, 11/2/3 on ESM2-150M while results in pretraining from scratch are mixed (6/8/2). Gains are seen for multiple models on 11 of 16 tasks, including stability, \b{eta}β\b{eta}-lactamase fitness, variant effect, intrinsic disorder, remote homology, enzyme classification, and SCOPe-40 fold retrieval. Tasks with more losses than wins are Fluorescence (TAPE) and Peptide-HLA Binding. All-position MLM+JEPA matches MLM-only overall but does not reproduce the masked-position gains. JEPA-only (no MLM) collapses in nearly every experiment. We conclude that JEPA, when combined with MLM, is competitive and can outperform pure MLM in pretraining and continued training, even under matched wall-clock budgets.

Artificial intelligence (AI) is increasingly central to understanding how the brain processes information. However, the integration of neuroscience and modern AI is bottlenecked by a fragmented software ecosystem. Current tools are siloed by recording modality and optimized for small-scale, in-memory workflows, limiting the use of massive, naturalistic datasets. Here, we introduce NeuralSet, a Python framework that efficiently unifies the processing of diverse neural recordings (including fMRI, M/EEG, and spikes) and complex experimental stimuli (such as text, audio, and video). By decoupling experimental metadata from lazy, memory-efficient data extraction, NeuralSet harmonizes standard neuroscientific preprocessing pipelines with pretrained deep learning embeddings. This approach provides a single PyTorch-ready interface that scales seamlessly from local prototyping to high-performance cluster execution. By eliminating manual data wrangling and ensuring full computational provenance, NeuralSet establishes a scalable, unified infrastructure for the next generation of neuro-AI research.

Adaptive cognition requires structured internal models of objects and their relations. Predictive neural networks are often proposed to learn such world models, but how these are instantiated and how they support prediction remain unclear. We investigate this in a minimal in-silico setting. A recurrent neural network samples tokens sequentially from 2D continuous token scenes and is trained to predict the upcoming token from the current input and a saccade-like displacement. On novel scenes, prediction accuracy improves across the sequence, indicating in-context learning. Decoding analyses reveal path integration and dynamic binding of token identity to position. Interventional analyses show that new bindings can be learned late in sequence and that out-of-distribution bindings can be learned as well. Together, these findings show how structured representations relying on flexible binding emerge to support prediction, offering a mechanistic account of sequential world modeling relevant to cognitive science.

Antimicrobial resistance (AMR) threatens global health. A promising and underexplored strategy to tackle this problem is sequential therapies exploiting collateral sensitivity (CS), whereby resistance to one drug increases sensitivity to another. Here, we develop a four-genotype stochastic birth-death model with two bacteriostatic antibiotics to identify switching periods that maximize bacterial extinction under subinhibitory concentrations. We show that extinction probability depends nonlinearly on switching period, with stepwise increases aligned to discrete switch events: fast sequential therapies are suboptimal as they do not allow for the evolution of resistance, a key ingredient in these therapies. A geometric distribution framework accurately predicts cumulative extinction probabilities, where the per-switch extinction probability rises with switching period. We further derive a heuristic approximation for the extinction probability based on times to fixation of single-resistant mutants. Sensitivity analyses reveal that strong reciprocal CS is required for this strategy to work, and we explore how increasing antibiotic doses and higher mutation rates modulate extinction in a nonmonotonic manner. Finally, we discuss how longer therapies maximize extinction but also cause higher resistance, leading to a Pareto front of optimal switching periods. Our results provide quantitative design principles for in vitro and clinical sequential antibiotic therapies, underscoring the potential of CS-guided regimens to suppress resistance evolution and eradicate infections.

Enteric methane measurement from ruminant livestock faces fundamental trade-offs between accuracy and operational feasibility. Existing methods quantify methane after eructation and atmospheric dilution, limiting temporal resolution and confounding biological signals with environmental variables. We present SCOUT (Smart Cannula-mounted Optical Unit for Trace-methane), the first autonomous system for continuous in-vivo monitoring of ruminal headspace methane concentrations. The system addresses a critical engineering barrier through closed-loop gas recirculation that maintains anaerobic ruminal conditions during persistent headspace sampling. SCOUT was deployed on cannulated Simmental heifers under contrasting dietary treatments. Headspace concentrations were 100 to 1000 times higher than concurrent ambient sniffer readings, providing substantially greater signal resolution for characterizing methane dynamics. High-frequency monitoring revealed behavior-production coupling previously inaccessible, including rapid concentration changes ($14.5 \pm 11.3k$ ppm) associated with postural transitions within 15-minute intervals. Cross-platform comparison with ambient sniffers showed scale-dependent correspondence between production and release measurements, with an optimal correlation (r = -0.564) at 40-minute averaging windows consistent with eructation cycles. These results demonstrate that the rumen headspace contains continuous, biologically interpretable methane signals that SCOUT can reliably access, establishing the measurement infrastructure necessary for developing concentration-to-flux models that would support precision phenotyping, emission proxy calibration, and mitigation strategy evaluation.

In general, the rates of infection and removal (whether through recovery or death) are nonlinear functions of the number of infected and susceptible individuals. One of the simplest models for the spread of infectious diseases is the SIR model, which categorizes individuals as susceptible, infectious, recovered or deceased. In this model, the infection rate, governing the transition from susceptible to infected individuals, is given by a linear function of both susceptible and infected populations. Similarly, the removal rate, representing the transition from infected to removed individuals, is a linear function of the number of infected individuals. While nonlinear infection and removal rates have been extensively studied in deterministic epidemiological models, analytic results for stochastic dynamics with general nonlinear rates remain limited. This work presents an analytic expression for the number of infected individuals considering nonlinear infection and removal rates. In particular, we examine how the number of infected individuals varies as cases emerge and obtain the expression accounting for the number of infected individuals at each moment. This work paves the way for new quantitative approaches to understanding infection dynamics.

Modeling the joint distribution of data samples and their properties allows to construct a single model for both data generation and property prediction, with synergistic benefits reaching beyond purely generative or predictive models. However, training joint models presents daunting architectural and optimization challenges. Here, we propose Hyformer, a transformer-based joint model that successfully blends the generative and predictive functionalities, using an alternating attention mechanism and a joint pre-training scheme. We show that Hyformer is simultaneously optimized for molecule generation and property prediction, while exhibiting synergistic benefits in conditional sampling, out-of-distribution property prediction and representation learning. Finally, we demonstrate the benefits of joint learning in a drug design use case of discovering novel antimicrobial~peptides.

Few studies have investigated the diagnostic utilities of biomarkers for predicting bacteremia among septic patients admitted to intensive care units (ICU). Therefore, this study evaluated the prediction power of laboratory biomarkers to utilize those markers with high performance to optimize the predictive model for bacteremia. This retrospective cross-sectional study was conducted at the ICU department of Gyeongsang National University Changwon Hospital in 2019. Adult patients qualifying SEPSIS-3 (increase in sequential organ failure score greater than or equal to 2) criteria with at least two sets of blood culture were selected. Collected data was initially analyzed independently to identify the significant predictors, which was then used to build the multivariable logistic regression (MLR) model. A total of 218 patients with 48 cases of true bacteremia were analyzed in this research. Both CRP and PCT showed a substantial area under the curve (AUC) value for discriminating bacteremia among septic patients (0.757 and 0.845, respectively). To further enhance the predictive accuracy, we combined PCT, bilirubin, neutrophil lymphocyte ratio (NLR), platelets, lactic acid, erythrocyte sedimentation rate (ESR), and Glasgow Coma Scale (GCS) score to build the predictive model with an AUC of 0.907 (95% CI, 0.843 to 0.956). In addition, a high association between bacteremia and mortality rate was discovered through the survival analysis (0.004). While PCT is certainly a useful index for distinguishing patients with and without bacteremia by itself, our MLR model indicates that the accuracy of bacteremia prediction substantially improves by the combined use of PCT, bilirubin, NLR, platelets, lactic acid, ESR, and GCS score.

The emergence of a hantavirus variant aboard a commercial cruise ship presents a significant public health concern. This study develops a discrete-time stochastic Susceptible-Exposed-Infectious-Recovered-Dead model to estimate transmission dynamics, hidden exposed infections, and outbreak risk among passengers and crew. Epidemiological parameters and latent disease states were inferred using an Ensemble Adjustment Kalman Filter calibrated to reported case data from WHO and ECDC situation reports. The estimated basic reproduction number was 2.76, with a 95\% confidence interval of 2.52-2.99, indicating substantial potential for sustained onboard transmission before strict quarantine measures. Simulations further suggest that several exposed individuals may remain unidentified during the early outbreak phase, creating a hidden reservoir that symptom-based surveillance alone may fail to detect. These findings highlight the importance of rapid surveillance, widespread testing, targeted quarantine, and active monitoring of exposed individuals in confined travel settings. The proposed modeling framework can support timely outbreak assessment and intervention planning for infectious-disease events in similarly dense and spatially constrained populations.

The success of machine learning in drug discovery hinges on learning the relationship between a chemical structure and its biological activity. While DNA-Encoded Library (DEL) technology can generate the massive datasets required for this task, its primary signal -- sequencing read counts -- is an indirect and often noisy proxy for true molecular binding affinity. To address the scarcity of public benchmarks for developing robust models that can overcome this data challenge, we introduce CA-DEL, a multi-dimensional public benchmark featuring screens against three homologous carbonic anhydrase isoforms. While recent benchmarks like KinDEL have introduced 3D poses for kinase targets, CA-DEL distinguishes itself by focusing on the selectivity challenge among homologous Carbonic Anhydrase isoforms (CAII, CAIX, CAXII). Unlike benchmarks relying solely on noisy enrichment scores, CA-DEL integrates a rigorous validation set of experimentally determined binding affinities ($K_i$) from ChEMBL, establishing a critical Sim-to-Real evaluation paradigm: training on noisy DEL screens and testing on high-fidelity biophysical data.

A central challenge in the origin of life is understanding how catalytic peptide-like polymers and information-bearing nucleic acid-like polymers emerged as an interde-pendent system. This study constructs a primordial cognitive model incorporating two internal Lotka-Volterra chemical oscillators to investigate, through simulation, whether a catalytic loop, primordial tRNAs, and nucleic acids that record and amplify them, can form through the interaction of polymers represented by binary (0/1) sequences. In this model, a mechanism was introduced where the synthesis of internal oscillations pro-vides a temporal bias for 0/1 selection during polymer elongation, while generated functional sequences are protected, recorded, and re-amplified. Simulation results demonstrated that the proposed cognitive model significantly outperformed a contrast model based on random 0/1 selection in terms of the establishment rate of catalytic loops, the accumulation of functional molecules, polymer elongation, and the reduction of Shannon entropy in sequence distribution. Furthermore, this superiority was generally maintained across sensitivity analyses, including batch calculations with different ran-dom seeds. While this study is a computational model based on abstract binary se-quences and simplified translation/replication rules rather than a direct reconstruction of life's origin, it provides a working hypothesis for the interdependent emergence of catalytic function and information retention by demonstrating that internal oscillations can bias sequence exploration within a framework linking autocatalytic networks, re-cording, and group selection. Future research must verify the generality and empirical validity of this framework by expanding monomer types, evolving into multi-oscillator systems, and establishing correspondences with compartmentalized experimental sys-tems.

The study of cultural evolution seeks to understand the processes by which behavioral variants are chosen in cultures over time, often as the result of large numbers of individual human choices. The selection of new popes, each of whom chooses a papal name -- typically reusing previous names in reference to previous popes -- is among the longest ongoing cultural processes taking place in a single human institution. Here, we use the record of papal names as a setting for long-term analysis of human cultural behavior. Although papal name choices are careful individual decisions, we find that the long-term sequence of papal names accords with predictions of a family of models developed in population genetics and stochastic processes -- Ewens sampling theory and the Chinese restaurant process -- which in the case of papal names amounts to randomly copying an existing name in proportion to its frequency, with the possibility of innovation of new names (mutations). Hence, despite the consideration that enters into choices of individual papal names, aggregate cultural behavior in a 2000-year old human process can potentially be described with simple laws. We discuss instances in which particular historical events might have caused temporary deviations from the random-copying model.

Functional connectivity (FC) derived from resting-state fMRI is widely used to characterize large-scale brain network alterations in neurological and psychiatric disorders. However, FC construction critically depends on the choice of brain atlas, and different parcellations may emphasize distinct organizational features, leading to heterogeneous and sometimes inconsistent representations. Existing multi-atlas approaches partially alleviate this issue but often fuse atlas-derived features or predictions at a relatively shallow level, while single-atlas disentanglement methods do not explicitly address cross-atlas heterogeneity. We propose Multi-Atlas Disentangled Connectivity LEarning (MADCLE), a multi-branch representation learning framework that jointly encodes FC matrices derived from different brain atlases. Rather than introducing a single explicitly shared latent variable across parcellations, MADCLE learns atlas-wise disease-related representations and encourages them to be cross-atlas consistent through distributional alignment. Meanwhile, covariate-related and atlas-dependent residual factors are modeled separately using covariate similarity supervision, atlas-specific reconstruction, and decorrelation constraints, thereby reducing the leakage of non-disease and parcellation-dependent information into the disease-related embeddings. Experiments on the ADNI and ADHD-200 datasets suggest that MADCLE achieves competitive or improved performance compared with single-atlas baselines, multi-atlas GNN/Transformer models, and recent multi-atlas consistency frameworks. These results support the potential value of structured disentanglement for FC-based disorder identification under heterogeneous parcellation schemes.

Our understanding of cell division control in bacteria still relies largely on interpreting correlations between phenomenological variables, with limited connection to the underlying molecular mechanisms. Here, we analytically solve a stochastic threshold-accumulation model in which a size-dependent divisor protein triggers division upon reaching a noisy, autocorrelated threshold, quantifying within a unified framework the combined effects of intrinsic and extrinsic noise and key mechanistic parameters such as protein reset and threshold memory. We show that incorporating these elements yields behavior far richer than the commonly assumed adder, spanning a continuum of division strategies from timer to sizer while modulating size fluctuations in a nontrivial fashion. Comparison with single-cell E. coli data shows that extrinsic noise and additional mechanistic ingredients are required to account for the observed size fluctuations. The adder emerges when threshold correlations balance protein reset, generalizing the hypothesis that full reset is necessary to maintain adder control. Our results establish a unified analytical framework linking stochastic molecular processes to emergent division laws, to be used in more complex bacterial cell-cycle models.

Trial-to-trial variability of neural responses has been linked to important aspects of neural computation and is essential for understanding how neuronal populations respond. While current overdispersion models treat each neuron's gain as independent of each other, this assumption fails to capture the network statistics of neuronal populations. As no existing model can capture overdispersed structured spiking gain-modulation across a neural population, network-level gain covariance remains largely unstudied. We thus present the Poisson matrix-normal latent variable (PMNLV) model, which extends single-neuron overdispersion to neural populations by placing a matrix-normal prior over the latent gain with a Kronecker-factored covariance. Spike counts are Poisson-distributed with a rate equal to the sum of a per-neuron stimulus tuning term and a matrix-normal gain, passed through a quadratic soft-rectifying link. We derive two complementary estimation algorithms: a variational EM (VEM) with a matrix-normal posterior that recovers dense Kronecker factors without structural assumptions, and a Kernel Tournament Method (KTM) that performs data-driven selection over a biologically motivated kernel dictionary and composite likelihood. On simulated data, both algorithms recover the inter-neuron and temporal covariance factors and accurate tuning curves. Applying VEM to Neuropixel recordings across four cortical regions of mouse visual hierarchy, we replicate a previous finding that single-neuron marginal variability changes little across cortical areas. We then show that shared population co-variability, invisible to scalar summaries e.g., the Fano factor, peaks in primary visual cortex and declines in higher visual areas. The PMNLV framework is applicable to any simultaneously recorded population where structured gain covariance is of scientific interest.

Protein sequence data from nature exhibits survivorship bias: we only observe data from those organisms that survive and reproduce, while non-functional protein mutations are eliminated by natural selection. Thus, predicting whether a protein sequence is functional often requires learning from positive examples alone. While positive-unlabeled (PU) learning frameworks offer a generic solution to this problem, existing PU methods ignore the evolutionary processes that shape sequence observability and cause survivorship bias. Consider a sequence that is one mutation away from a commonly-observed protein variant in a well-surveilled organism. If the sequence were functional, it would likely be observed. If it is not observed, this suggests non-functionality. In contrast, sequences that are unlikely to arise through mutation may be missing simply because they never arose. Thus, these two kinds of missing sequences should be treated differently when training models. In this work, we propose Evo-PU, a PU learning framework that uses a scientific understanding of nucleotide mutation to model survivorship bias for well-surveilled single-organism sequence data. On three prediction tasks using single-organism uniform-coverage surveillance data -- predicting results from held-out influenza and respiratory syncytial virus (RSV) mutagenesis studies, and predicting future SARS-CoV-2 variants -- Evo-PU outperforms standard PU learning, one-class classification (OCC), and protein language models (PLMs). On prediction tasks from multi-organism ProteinGym datasets with more heterogeneous surveillance coverage, we identify opportunities to generalize our approach.

Robust genotype-to-phenotype (G2P) prediction is essential for accelerating breeding decisions and genetic gain. However, it remains challenging to measure complex traits under variable field conditions and across years. In this study, we propose a linear-Transformer approach, LiT-G2P (Linear-Transformer Genotype-to-Phenotype), an automated predictive framework that integrates additive genetic variance effects with Transformer-based nonlinear interactions using genome-wide single-nucleotide polymorphisms (SNPs) data. We evaluated LiT-G2P on a panel of diverse grape accessions, genotyped with SNP markers and measured for phenotypes across two consecutive years. Target phenotypic traits include leaf hair density and trichome density of grapevines. Across both single-year and cross-year testing scenarios, LiT-G2P consistently improves prediction performance compared with baseline models. For hair density, LiT-G2P achieves the lowest error in both single-year and cross-year evaluations, with RMSEs of 0.469 and 0.454, respectively, while maintaining strong tolerance accuracies of 79.2% and 74.6%, respectively. For trichome density, LiT-G2P also presents the best overall G2P performance. In addition, we extract model-prioritized SNPs from attention weights and apply genotype-stratified analysis to provide interpretable candidate marker for downstream validation. These results demonstrate that integrating stable additive effects with learned interaction patterns can enhance cross-year robustness and support practical SNP-based predictive modeling for genomic selection.

Interpreting transcriptomic data is one of the most common analytical tasks in modern biology. Yet most current models either consume expression profiles without producing natural-language biological explanations, or reason in language without direct access to quantitative omics measurements. We introduce OmicsLM, a multimodal LLM that connects quantitative omics profiles with natural-language biological tasks. OmicsLM represents each transcriptomic profile as a compact continuous representation within the LLM context. This interface preserves quantitative expression signal while allowing natural-language instructions, explicit gene mentions, and multiple interleaved biological samples to be processed together in one model context. We train OmicsLM on more than 5.5 million instruction-following examples spanning over 70 task types, combining continuous transcriptomic inputs, experimental data rendered through diverse language templates, and free-text biological knowledge and question-answering data. This mixture covers cell type annotation, perturbation prediction, clinical prediction, pathway reasoning, and open-ended biological question answering. Existing benchmarks evaluate either profile-level prediction or text-only biological QA, leaving language-guided, multi-sample reasoning over real expression profiles unmeasured. To close this gap, we introduce GEO-OmicsQA, a benchmark for multi-sample biological question answering built from real Gene Expression Omnibus (GEO) studies. We demonstrate that OmicsLM can use expression profiles directly and perform comparably to specialized omics models on profile-level tasks, while outperforming both omics-specialized models and general LLMs on language-guided biological reasoning over expression data.

Neuroimaging does not observe causal variables directly: hemodynamics and volume conduction distort signals so that statistical dependence need not reflect latent neural influence. Before estimating graphs, one must specify under what assumptions delayed directed structure can be studied from such indirect observations. We formalize a conditional setting - recoverable inversion under modality physics together with nonstationary latent dynamics - and derive an inversion-error propagation bound under explicit assumptions. Building on this framing, we propose INCAMA (INdirect CAusal MAmba): physics-aware inversion coupled with a delay-aware Mamba encoder that uses mechanism shifts as informative variation for directed graph scoring. We use controlled simulations for quantitative validation and HCP motor-task fMRI as a zero-shot external transfer check based on anatomical and task-network consistency. Across TVB simulations, INCAMA improves directed-structure recovery by 2-3x in F1 over observation-space and two-stage baselines, and on HCP motor-task fMRI it produces sparse directed estimates concentrated in canonical visuo-motor pathways.

Molecular function is largely determined by structure. Accurately aligning molecular structure with natural language is therefore essential for enabling large language models (LLMs) to reason about downstream chemical tasks. However, the substantial cost of human annotation makes it infeasible to construct large-scale, high-quality datasets of structure-grounded descriptions. In this work, we propose a fully automated annotation framework for generating precise molecular descriptions that preserve complete structural details at scale. Our approach builds upon and extends a rule-based chemical nomenclature parser to interpret IUPAC names and construct enriched, structural XML metadata that explicitly encodes molecular structure. This metadata is then used to guide LLMs in producing accurate natural-language descriptions. Using this framework, we curate a large-scale dataset of approximately $163$k molecule--description pairs. A rigorous validation protocol combining LLM-based and expert human evaluation on a subset of $2,000$ molecules demonstrates a high description precision of $98.6$%. The proposed annotation framework is readily beneficial to broader chemical tasks that rely on structural descriptions, with the resulting dataset providing a reliable foundation for molecule--language alignment. The source code and dataset are hosted at https://github.com/TheLuoFengLab/MolLangData and https://huggingface.co/datasets/ChemFM/MolLangData, respectively.

Information can evolve as a physical consequence of non-equilibrium dynamics, even in the absence of genes, replication, or predefined fitness functions. We present Stability-Driven Assembly (SDA), a framework in which stochastic assembly combined with differential persistence biases populations toward longer-lived motifs. Assemblies that persist longer become more frequent and are therefore more likely to participate in subsequent interactions, generating feedback that reshapes the population distribution and implements fitness-proportional sampling, realizing evolution as a natural, emergent genetic algorithm (SDA/GA) driven solely by stability. We apply SDA/GA to chemical symbol space using SMILES fragments with recombination, mutation, and a heuristic stability function. Simulations show hallmark features of evolutionary search, including scaffold-level dominance, sustained novelty, and entropy reduction, yielding open-ended dynamics absent from equilibrium models with fixed transition rates. These results motivate an evolutionary ladder hypothesis where persistence-driven selection precedes genetic replication.

Biological systems can form complex three-dimensional structures through the collective behavior of agents that share a common update rule and operate without central control. How such distributed control gives rise to precise global patterns remains a central question not only in developmental biology but also in distributed robotics, programmable matter, and multi-agent learning. Here, we introduce DiffeoMorph, an end-to-end differentiable framework for learning a morphogenesis protocol that guides a population of agents to morph into a target 3D shape. Each agent updates its position and internal state using an SE(3)-equivariant graph neural network, based on its own internal state and signals received from other agents. To train this system, we introduce a new shape-matching loss based on 3D Zernike polynomials, which compares the predicted and target shapes as continuous spatial distributions, not as discrete point clouds, and is invariant to agent ordering, number of agents, and global orientation. To achieve rotation invariance while preserving reflection sensitivity, we include an alignment step that optimally rotates the predicted Zernike spectrum to match the target before computing the loss. We perform benchmarking to establish the advantages of our shape-matching loss over other standard distance metrics for shape comparison tasks. We then demonstrate that DiffeoMorph can form a range of complex shapes from minimally patterned initial conditions. DiffeoMorph provides a general framework for learning distributed control strategies for morphogenesis, swarm robotics, and programmable self-assembly.

Generating property-optimized mRNA sequences is central to applications such as vaccine design and protein replacement therapy, but remains challenging due to limited data, complex sequence-function relationships, and the narrow space of biologically viable sequences. Generative methods that drift away from the data manifold can yield sequences that fail to fold, translate poorly, or are otherwise nonfunctional. We present RNAGenScape, a property-guided manifold Langevin dynamics framework for mRNA sequence generation that operates directly on a learned manifold of real data. By performing iterative local optimization constrained to this manifold, RNAGenScape preserves biological viability, accesses reliable guidance, and avoids excursions into nonfunctional regions of the ambient sequence space. The framework integrates three components: (1) an autoencoder jointly trained with a property predictor to learn a property-organized latent manifold, (2) a denoising autoencoder that projects updates back onto the manifold, and (3) a property-guided Langevin dynamics procedure that performs optimization along the manifold. Across three real-world mRNA datasets spanning two orders of magnitude in size, RNAGenScape increases median property gain by up to 148% and success rate by up to 30% while ensuring biological viability of generated sequences, and achieves competitive inference efficiency relative to existing generative approaches.

Neural recordings exhibit a distinctive form of heterogeneity rooted in differences in cell types, intrinsic circuit dynamics, and stochastic stimulus-response variability that goes beyond ordinary dataset variability, mixing statistically regular neurons with highly stochastic, stimulus-contingent ones within the same dataset. This heterogeneity poses a challenge for self-supervised learning (SSL) -- learnable statistical regularity -- thereby destabilizing representation learning and limiting reliable scaling. We introduce POYO-CAP (Cell-pattern Aware Pretraining), a biologically grounded hybrid pretraining strategy that first trains with masked reconstruction plus lightweight auxiliary supervision on statistically regular neurons -- identified via skewness and kurtosis -- and then fine-tunes on more stochastic populations. On the Allen Brain Observatory dataset, this curriculum yields 12--13\% relative improvements over from-scratch training and enables smooth, monotonic scaling with model size, whereas baselines trained on mixed populations plateau or destabilize. By making statistical predictability an explicit data-selection criterion, POYO-CAP turns neural heterogeneity into a scalable learning advantage for robust neural decoding.