Learning systems must balance generalization across experiences with discrimination of task-relevant details. Effective learning therefore requires representations that support both. Online latent-cause models support incremental inference but assume flat partitions, whereas hierarchical Bayesian models capture multilevel structure but typically require offline inference. We introduce the Hierarchical Online Learning of Multiscale Experience Structure (HOLMES) model, a computational framework for hierarchical latent structure learning through online inference. HOLMES combines a variation on the nested Chinese Restaurant Process prior with sequential Monte Carlo inference to perform tractable trial-by-trial inference over hierarchical latent representations without explicit supervision over the latent structure. In simulations, HOLMES matched the predictive performance of flat models while learning more compact representations that supported one-shot transfer to higher-level latent categories. In a context-dependent task with nested temporal structure, HOLMES also improved outcome prediction relative to flat models. These results provide a tractable computational framework for discovering hierarchical structure in sequential data.

We present BSTModelKit.jl, an open-source Julia package for constructing, solving, and analyzing Biochemical Systems Theory (BST) models of biochemical networks. The package implements S-system representations, a canonical power-law formalism for modeling metabolic and regulatory networks. BSTModelKit.jl provides a declarative model specification format, dynamic simulation via ordinary differential equation (ODE) integration, steady-state computation, and global sensitivity analysis using the Morris and Sobol methods. The package leverages the Julia scientific computing ecosystem, in particular the SciML suite of differential equation solvers, to provide efficient and flexible model analysis tools. We describe the mathematical formulation, software design, and demonstrate the package capabilities with illustrative examples.

The SIR model is the cornerstone model for mathematical epidemiology, explaining key epidemic features such as the second-order transition between disease-free and epidemic states, the initial exponential growth of outbreaks or the short-term benefits of control measures. Nonetheless, the classical SIR model assumes that pathogen traits remain fixed, thus neglecting viral evolution. Here we propose a minimal extension of the SIR model, allowing infectiousness to evolve. We show that such evolution can cause superexponential early growth of outbreaks, create abrupt epidemic transitions, and undermine the effectiveness of control policies, as lifting interventions too early can lead to worse epidemic scenarios than no action. We derive analytical expressions for the critical mutation rate and intervention time governing this behavior, and identify a strong asymmetry between control strategies: while shortening the infectious period hinders transmission without suppressing viral evolution, lowering transmission both reduces cases and slows down viral evolution.

At present, there are no easily understood explainable artificial intelligence (AI) methods for discrete token inputs, like text. Most explainable AI techniques do not extend well to token sequences, where both local and global features matter, because state-of-the-art models, like transformers, tend to focus on global connections. Therefore, existing explainable AI algorithms fail by (i) identifying disparate tokens of importance, or (ii) assigning a large number of tokens a low value of importance. This method for explainable AI for tokens-based classifiers generalizes a mask-based explainable AI algorithm for images. It starts with an Explainer neural network that is trained to create masks to hide information not relevant for classification. Then, the Hadamard product of the mask and the continuous values of the classifier's embedding layer is taken and passed through the classifier, changing the magnitude of the embedding vector but keeping the orientation unchanged. The Explainer is trained for a taxonomic classifier for nucleotide sequences and it is shown that the masked segments are less relevant to classification than the unmasked ones. This method focused on the importance the token as a whole (i.e., a segment of the input sequence), producing a human-readable explanation.

Lipophilicity (logP) prediction remains central to drug discovery, yet linear regression models for this task frequently violate statistical assumptions in ways that invalidate their reported performance metrics. We analyzed 426,850 bioactive molecules from a rigorously curated intersection of PubChem, ChEMBL, and eMolecules databases, revealing severe heteroskedasticity in linear models predicting computed logP values (XLOGP3): residual variance increases 4.2-fold in lipophilic regions (logP greater than 5) compared to balanced regions (logP 2 to 4). Classical remediation strategies (Weighted Least Squares and Box-Cox transformation) failed to resolve this violation (Breusch-Pagan p-value less than 0.0001 for all variants). Tree-based ensemble methods (Random Forest R-squared of 0.764, XGBoost R-squared of 0.765) proved inherently robust to heteroskedasticity while delivering superior predictive performance. SHAP analysis resolved a critical multicollinearity paradox: despite a weak bivariate correlation of 0.146, molecular weight emerged as the single most important predictor (mean absolute SHAP value of 0.573), with its effect suppressed in simple correlations by confounding with topological polar surface area (TPSA). These findings demonstrate that standard linear models face fundamental challenges for computed lipophilicity prediction and provide a principled framework for interpreting ensemble models in QSAR applications.

Histopathological analysis is a cornerstone of cancer diagnosis, with Hematoxylin and Eosin (H&E) staining routinely acquired for every patient to visualize cell morphology and tissue architecture. On the other hand, multiplex immunofluorescence (mIF) enables more precise cell type identification via proteomic markers, but has yet to achieve widespread clinical adoption due to cost and logistical constraints. To bridge this gap, we introduce MIPHEI (Multiplex Immunofluorescence Prediction from H&E Images), a U-Net-inspired architecture that leverages a ViT pathology foundation model as encoder to predict mIF signals from H&E images using rich pretrained representations. MIPHEI targets a comprehensive panel of markers spanning nuclear content, immune lineages (T cells, B cells, myeloid), epithelium, stroma, vasculature, and proliferation. We train our model using the publicly available OrionCRC dataset of restained H&E and mIF images from colorectal cancer tissue, and validate it on five independent datasets: HEMIT, PathoCell, IMMUcan, Lizard and PanNuke. On OrionCRC test set, MIPHEI achieves accurate cell-type classification from H&E alone, with F1 scores of 0.93 for Pan-CK, 0.83 for alpha-SMA, 0.68 for CD3e, 0.36 for CD20, and 0.28 for CD68, substantially outperforming both a state-of-the-art baseline and a random classifier for most markers. Our results indicate that, for some molecular markers, our model captures the complex relationships between nuclear morphologies in their tissue context, as visible in H&E images and molecular markers defining specific cell types. MIPHEI offers a promising step toward enabling cell-type-aware analysis of large-scale H&E datasets, in view of uncovering relationships between spatial cellular organization and patient outcomes.

Examining Drug-Drug Interactions (DDIs) is a pivotal element in the process of drug development. DDIs occur when one drug's properties are affected by the inclusion of other drugs. Detecting favorable DDIs has the potential to pave the way for creating and advancing innovative medications applicable in practical settings. However, existing DDI prediction models continue to face challenges related to generalization in extreme cases, robust feature extraction, and real-life application possibilities. We aim to address these challenges by leveraging the effectiveness of context-aware deep graph learning by introducing a novel framework named CADGL. Based on a customized variational graph autoencoder (VGAE), we capture critical structural and physio-chemical information using two context preprocessors for feature extraction from two different perspectives: local neighborhood and molecular context, in a heterogeneous graphical structure. Our customized VGAE consists of a graph encoder, a latent information encoder, and an MLP decoder. CADGL surpasses other state-of-the-art DDI prediction models, excelling in predicting clinically valuable novel DDIs, supported by rigorous case studies. CADGL is vailable at: https://github.com/azminewasi/cadgl

Subcellular localization is a crucial biological task for drug target identification and function annotation. Although it has been biologically realized that subcellular localization is closely associated with protein structure, no existing dataset offers comprehensive 3D structural information with detailed subcellular localization annotations, thus severely hindering the application of promising structure-based models on this task. To address this gap, we introduce a new benchmark called $\mathbf{CAPSUL}$, a $\mathbf{C}$omprehensive hum$\mathbf{A}$n $\mathbf{P}$rotein benchmark for $\mathbf{SU}$bcellular $\mathbf{L}$ocalization. It features a dataset that integrates diverse 3D structural representations with fine-grained subcellular localization annotations carefully curated by domain experts. We evaluate this benchmark using a variety of state-of-the-art sequence-based and structure-based models, showcasing the importance of involving structural features in this task. Furthermore, we explore reweighting and single-label classification strategies to facilitate future investigation on structure-based methods for this task. Lastly, we showcase the powerful interpretability of structure-based methods through a case study on the Golgi apparatus, where we discover a decisive localization pattern $α$-helix from attention mechanisms, demonstrating the potential for bridging the gap with intuitive biological interpretability and paving the way for data-driven discoveries in cell biology.

Inferring the connectivity of neural circuits from incomplete observations is a fundamental challenge in neuroscience. We present a covariance-based method for estimating the weight matrix of a recurrent neural network from sparse, partial measurements across multiple recording sessions. By accumulating pairwise covariance estimates across sessions where different subsets of neurons are observed, we reconstruct the full connectivity matrix without requiring simultaneous recording of all neurons. A Granger-causality refinement step enforces biological constraints via projected gradient descent. Through systematic experiments on synthetic networks modeling small brain circuits, we characterize a fundamental control-estimation tradeoff: stimulation aids identifiability but disrupts intrinsic dynamics, with the optimal level depending on measurement density. We discover that the ``incorrect'' linear approximation acts as implicit regularization -- outperforming the oracle estimator with known nonlinearity at all operating regimes -- and provide an exact characterization via the Stein--Price identity.

Spatial transcriptomics (ST) profiles gene expression across a tissue section while preserving the spatial coordinates. Because current ST technologies typically profile two-dimensional tissue slices, integrating and aligning slices from different regions of the same three-dimensional tissue or from samples under different conditions enables analyses that reveal 3D organization and condition-associated spatial patterns. Two major challenges remain. First, interpretable and flexible control over spatial distortion is needed because rigid transformations can be overly restrictive, whereas highly deformable mappings may arbitrarily distort spatial proximity. Second, biologically plausible matching is also needed, especially when the slices overlap partially. Here, we introduce RAFT-UP, a tool for robust ST alignment that provides explicit control over spatial distance preservation through a fused supervised Gromov-Wasserstein (FsGW) optimal transport framework. FsGW combines expression and spatial information, incorporates spot-wise constraints to discourage biologically implausible matches, and enforces a pairwise distance-consistency constraint that prevents mapping two pairs of spots when their spatial distances differ beyond a specified tolerance. We demonstrate that RAFT-UP accurately aligns slices from different regions of the same tissue and slices from different samples. Benchmarking shows that RAFT-UP improves spatial distance preservation while achieving spot label matching accuracy comparable to state-of-the-art methods. Finally, we demonstrate RAFT-UP on two spatially constrained downstream applications, including spatiotemporal mapping of developing mouse midbrain and comparative cross-slice analysis of cell-cell communication. RAFT-UP is available as open-source software.

Background Genetic parameters of feeding behaviours traits from electronic feeding stations in relation to feed efficiency have been widely explored. However, genetic determinism of the circadian rhythm of feed intake throughout the fattening phase in group-housed growing pigs fed ad libitum has never been investigated, despite the well-known relationships between animals' circadian rhythms and the optimization of their metabolism. The objective of this study was to (i) propose three new traits derived from time-frequency approach applied to electronic feeding data from 2,297 Large White pigs that reflect the consistency of circadian feed intake rhythm throughout fattening (so called DayCR) and the precocity of its establishment (so called IndexCR and gCR), and then to (ii) estimate the heritability of those traits and their genetic correlations with residual feed intake using a multiple trait model. Results Results highlighted moderate heritability estimates for the three circadian traits (range h2: [0.24; 0.35]) and high heritability for residual feed intake (0.41). High genetic correlations (range of absolute values: [0.87; 0.98]) among circadian traits suggested that pigs exhibiting a 24-hour periodicity in feed intake on most days of fattening, particularly during the final fattening period, establish their circadian rhythm earlier than the other pigs. The low (range of absolute values: [0.18; 0.27]) but favourable genetic correlations between residual feed intake and circadian traits revealed that animals with a consistent and early 24-hour periodicity of feed intake also tend to be more feed efficient. Conclusions This study proposed to apply time-frequency analysis on longitudinal feeding data to detect 24-hour periodicities in the hourly feed intake pattern across days throughout fattening in growing-pigs. Results suggested that part of the variability observed in the establishment of circadian rhythm is genetically driven, further supporting the feasibility of genetic selection on circadian traits. Considering the well-established biological mechanisms underlying circadian feeding rhythm, selecting animals for their ability to exhibit an early and consistent 24-hour periodicity of feed intake could promote metabolic homeostasis, thereby enhancing animal performance and resilience.

Replica exchange (REX) is one of the most widely used enhanced sampling methodologies, yet its efficiency is limited by the requirement for a large number of intermediate temperature replicas. Here we present Generative Replica Exchange (GREX), which integrates deep generative models into the REX framework to eliminate this temperature ladder. Drawing inspiration from reservoir replica exchange (res-REX), GREX utilizes trained normalizing flows to generate high-temperature configurations on demand and map them directly to the target distribution using the potential energy as a constraint, without requiring target-temperature training data. This approach reduces production simulations to a single replica at the target temperature while maintaining thermodynamic rigor through Metropolis exchange acceptance. We validate GREX on three benchmark systems of increasing complexity, highlighting its superior efficiency and practical applicability for molecular simulations.

While explainable AI (XAI) is often heralded as a means to enhance transparency and trustworthiness in closed-loop neurotechnology for psychiatric and neurological conditions, its real-world prevalence remains low. Moreover, empirical evidence suggests that the type of explanations provided by current XAI methods often fails to align with clinicians' end-user needs. In this viewpoint, we argue that clinically meaningful explainability (CME) is essential for AI-enabled closed-loop medical neurotechnology and must be addressed from an ethical, technical, and clinical perspective. Instead of exhaustive technical detail, clinicians prioritize clinically relevant, actionable explanations, such as clear representations of input-output relationships and feature importance. Full technical transparency, although theoretically desirable, often proves irrelevant or even overwhelming in practice, as it may lead to informational overload. Therefore, we advocate for CME in the neurotechnology domain: prioritizing actionable clarity over technical completeness and designing interface visualizations that intuitively map AI outputs and key features into clinically meaningful formats. To this end, we introduce a reference architecture called NeuroXplain, which translates CME into actionable technical design recommendations for any future neurostimulation device. Our aim is to inform stakeholders working in neurotechnology and regulatory framework development to ensure that explainability fulfills the right needs for the right stakeholders and ultimately leads to better patient treatment and care.

Virtual cell models aim to enable in silico experimentation by predicting how cells respond to genetic, chemical, or cytokine perturbations from single-cell measurements. In practice, however, large-scale perturbation prediction remains constrained by three coupled bottlenecks: inefficient training and inference pipelines, unstable modeling in high-dimensional sparse expression space, and evaluation protocols that overemphasize reconstruction-like accuracy while underestimating biological fidelity. In this work we present a specialized large-scale foundation model SCALE for virtual cell perturbation prediction that addresses the above limitations jointly. First, we build a BioNeMo-based training and inference framework that substantially improves data throughput, distributed scalability, and deployment efficiency, yielding 12.51* speedup on pretrain and 1.29* on inference over the prior SOTA pipeline under matched system settings. Second, we formulate perturbation prediction as conditional transport and implement it with a set-aware flow architecture that couples LLaMA-based cellular encoding with endpoint-oriented supervision. This design yields more stable training and stronger recovery of perturbation effects. Third, we evaluate the model on Tahoe-100M using a rigorous cell-level protocol centered on biologically meaningful metrics rather than reconstruction alone. On this benchmark, our model improves PDCorr by 12.02% and DE Overlap by 10.66% over STATE. Together, these results suggest that advancing virtual cells requires not only better generative objectives, but also the co-design of scalable infrastructure, stable transport modeling, and biologically faithful evaluation.

Electroencephalography (EEG) foundation models hold significant promise for universal Brain-Computer Interfaces (BCIs). However, existing approaches often rely on end-to-end fine-tuning and exhibit limited efficacy under frozen-probing protocols, lacking the intrinsic universality required for broad generalization. This limitation stems from adapting general-purpose sequence architectures that overlook the biophysical and dynamical principles of neural activity. To bridge this gap, we propose DeeperBrain, a neuro-grounded foundation model integrating domain-specific inductive biases into its model design and learning objectives. Architecturally, DeeperBrain incorporates a volume conduction-aware channel encoding to model spatial mixing via 3D geometry, and a neurodynamics-aware temporal encoding capturing slow adaptations using oscillatory and exponential bases. For pretraining, we introduce a dual-objective strategy combining Masked EEG Reconstruction (MER) for local fidelity and Neurodynamics Statistics Prediction (NSP). NSP enforces alignment with macroscopic brain states by predicting interpretable order parameters, including spectral power, functional connectivity, cross-frequency coupling, and dynamic complexity. Extensive experiments demonstrate that DeeperBrain achieves state-of-the-art or highly competitive performance under end-to-end fine-tuning. Crucially, it maintains superior efficacy under a rigorous frozen-probing protocol, verifying that embedding neuroscientific first principles endows learned representations with the intrinsic universality essential for universal BCI. The code will be publicly available.

The evolution of natural languages poses a riddle to any theoretical perspective based on efficiency considerations. If languages are already optimally effective means of organization and communication of thought, why do they change? And if they are driven to become optimally effective in the future, why do they change so slowly, and why do they diversify, rather than converge towards an optimum? We look here at the hypothesis that disorder, rather than efficiency, may play a dominant role. Most traditional approaches to study diachronic language dynamics emphasize lexical data, but it would seem that a crucial contribution to the effectiveness of a thought-coding device is given by its core generative structure, i.e., its syntax. Based on the reduction of syntax to a set of binary syntactic parameters, we introduce here a model of natural language change in which diachronic dynamics stem from disordered interactions between/among parameters, even in the idealized limit of identical external inputs. We show in which region of phase space such dynamics show the glassy features that are observed in natural language across time. In particular, binary syntactic vectors remain trapped in glassy metastable (ie, tendentially stable) states when the degree of asymmetry in the disordered interactions is below a critical value, consistent with studies of spin glasses with asymmetric interactions. We further show that an added Hopfield-type memory term, would indeed, if strong enough, stabilize syntactic configurations, but losing their multiplicity. Finally, using a notion of linguistic distance in syntactic state space we show that a phylogenetic signal may remain among related languages, despite their gradually divergent syntax, exactly as recently pointed out for real-world languages. These statistical results appear to generalize beyond the 94 parameters across 58 languages used here.

We explore a functionalist approach to emotion by employing an ansatz -- an initial set of assumptions -- that a hypothetical concept generation model incorporates unproven but biologically plausible traits. From these traits, we mathematically construct a theoretical reinforcement learning framework grounded in functionalist principles and examine how the resulting utility function aligns with emotional valence in biological systems. Our focus is on structuring the functionalist perspective through a conceptual network, particularly emphasizing the construction of the utility function, not to provide an exhaustive explanation of emotions. The primary emphasis is not of planning or action execution, but such factors are addressed when pertinent. Finally, we apply the framework to psychological phenomena such as humor, psychopathy, and advertising, demonstrating its breadth of explanatory power.

Populations of cells regulate gene expression in response to external signals, but their ability to make reliable collective decisions is limited by both intrinsic noise in molecular signaling and variability between individual cells. In this work, we use optogenetic control of the canonical Wnt pathway as an example to study how reliably information about an external signal is transmitted to a population of cells, and determine an optimal encoding strategy to maximize information transmission from Wnt signals to gene expression. We find that it is possible to reach an information capacity beyond 1 bit only through an appropriate, discrete encoding of signals: using either no Wnt, a short Wnt pulse, or a sustained Wnt signal. By averaging over an increasing number of outputs, we systematically vary the effective noise in the pathway. As the effective noise decreases, the optimal encoding comprises more discrete input signals. These signals do not need to be fine-tuned to achieve near-optimal information transmission. The optimal code transitions into a continuous code in the small-noise limit, which can be shown to be consistent with the Jeffreys prior. We visualize the performance of different signal encodings using decoding maps. Our results suggest optogenetic Wnt signaling allows for regulatory control beyond a simple binary switch, and provides a framework to apply ideas from information processing to single-cell in vitro experiments.

Human cognition is supported by brain structural connectivity wherein weak connectivity with weights several orders of magnitude smaller than those of strong connectivity, has been treated as noise and ignored from analysis over a long time. We here propose that weak connectivity plays roles to cognitive abilities by nonlinearly amplifying its small weights. Using the HCP dataset (n=999) and multiple tractography algorithms, we found that weak connectivity involves high individual variability and contributes to predictions of general cognitive ability and memory, and it is also critical for brain functional connectivity simulation and structure-function coupling. Importantly, we fused two post-tractography filtering methods to generate more reliable connectivity that preserves weak links and outperforms conventional thresholding. At the network level, we showed that weak connectivity expands the operational capacity of brain networks to enhance both global integration and fine-grained segregation, thereby supporting a functional balance essential for diverse cognitive abilities. Finally, we identified a specific type of weak connectivity mainly linking visual/motor to limbic areas with negative gene co-expression, which has a disproportionately large functional impact. Our findings demonstrate groundbreaking evidence of underestimated but crucial role of weak connectivity in human cognition, providing a refined approach to reliably reveal brain structural connectivity.

In this article, I put forward the idea that the neoplastic process (NP) has deep evolutionary roots and make specific predictions about the connection between cancer and the formation of the first embryo, which allowed for the evolutionary radiation of metazoans. My main hypothesis is that the NP is at the heart of cellular mechanisms responsible for animal morphogenesis and, given its embryological basis, also at the center of animal evolution. It is thus understood that NP-associated mechanisms are deeply rooted in evolutionary history and tied to the formation of the first animal embryo. In my consideration of these arguments, I expound on how cancer biology is perfectly intertwined with evolutionary biology. I describe essential cellular components of unicellular holozoans that served as a basis for the formation of the neoplastic functional module (NFM) and its subsequent exaptation, which brought forth two great biophysical revolutions within the first embryo. Finally, I examine the role of Physics in the modeling of the NFM and its contribution to morphogenesis to reveal the totipotency of the zygote.

The decisive role of Embryology in understanding the evolution of animal forms is founded and deeply rooted in the history of science. It is recognized that the emergence of multicellularity would not have been possible without the formation of the first embryo. We speculate that biophysical phenomena and the surrounding environment of the Ediacaran ocean were instrumental in co-opting a neoplastic functional module (NFM) within the nucleus of the first zygote. Thus, the neoplastic process, understood here as a biological phenomenon with profound embryologic implications, served as the evolutionary engine that favored the formation of the first embryo and cancerous diseases and allowed to coherently create and recreate body shapes in different animal groups during evolution. In this article, we provide a deep reflection on the Physics of the first embryogenesis and its contribution to the exaptation of additional NFM components, such as the extracellular matrix. Knowledge of NFM components, structure, dynamics, and origin advances our understanding of the numerous possibilities and different innovations that embryos have undergone to create animal forms via Neoplasia during evolutionary radiation. The developmental pathways of Neoplasia have their origins in ctenophores and were consolidated in mammals and other apical groups.

To make nature mathematically tractable, the scientific model of the world omits qualia--colors, sounds, tastes, sensations--leaving only what admits of numerical characterization. The "hard problem" of consciousness--the enigma of why and how physical processing gives rise to felt experience--remains unsolved. The Quantification Horizon Theory of Consciousness (QHT) proposes that this enigma reflects a structural limitation of mathematical description: quantitative models capture only quantifiable features of reality; qualia are left out. Yet despite this limitation, QHT argues that such models can account for the unquantifiable--not by explaining it, but by registering its presence, in the form of a signpost. There are specific features of information geometry--compression singularities--that intuitively correspond to the hallmark properties of consciousness and could serve as precisely such signposts. QHT proposes that these singularities mark a quantification horizon--a boundary beyond which quantitative description cannot reach. On this proposal, qualia lie beyond the horizon. From this basis, the theory derives ineffability, privacy, and subjectivity as structural consequences and proposes structural accounts of unity and causal efficacy. The theory proposes substrate-independent dynamical criteria for determining which systems are plausible candidates for consciousness, avoids panpsychism, makes testable predictions, and offers concrete implications for artificial intelligence and artificial consciousness.