Quantifying the neural signatures of consciousness remains a major challenge in neuroscience and AI. Although many theories link consciousness to rich, multiscale, and flexible neural organisation, robust quantitative measures are still lacking. This paper presents a theory-neutral framework that characterises consciousness-related dynamics through three properties: hierarchical integration (H), cross-frequency complexity (D), and metastability (M). Candidate subsystems are identified using predictive information, temporal complexity, and state-space exploration to distinguish structured from unstructured activity. We provide mathematical definitions for all components and implement the framework in a generative model of synthetic EEG, simulating nine brain states ranging from psychedelic and wakeful to dreaming, non-REM sleep, minimally conscious, anaesthetised, and seizure-like regimes. Across single trials and Monte Carlo ensembles, the composite index reliably separates high-consciousness from impaired or non-conscious states. We further validate the framework using real EEG from the Sleep-EDF dataset alongside matched synthetic EEG designed to reproduce state-dependent oscillatory structure. Across Wake, N2, and REM sleep, synthetic data recapitulate the empirical ordering and magnitude of the index, indicating that the index captures stable and biologically meaningful distinctions. This approach provides a principled and empirically grounded tool for quantifying consciousness-related neural organisation with potential applications to both biological and artificial systems.

Biological intelligence emerges from substrates that are slow, noisy, and energetically constrained, yet it performs rapid and coherent inference in open-ended environments. Classical computational theories, built around vector-space transformations and instantaneous error minimization, struggle to reconcile the slow timescale of synaptic plasticity with the fast timescale of perceptual synthesis. We propose a unifying framework based on algebraic topology, the Homological Brain, in which neural computation is understood as the construction and navigation of topological structure. Central to this view is the Parity Principle, a homological partition between even-dimensional scaffolds encoding stable content ($\Phi$) and odd-dimensional flows encoding dynamic context ($\Psi$). Transient contextual flows are resolved through a three-stage topological trinity transformation: Search (open-chain exploration), Closure (topological cycle formation), and Condensation (collapse of validated flows into new scaffold). This process converts high-complexity recursive search (formally modeled by Savitch's Theorem in NPSPACE) into low-complexity navigation over a learned manifold (analogous to memoized Dynamic Programming in P). In this framework, topological condensation is the mechanism that transforms a ``search problem'' into a ``navigation task'', allowing the brain to amortize past inference and achieve rapid perceptual integration. This perspective unifies the Wake-Sleep cycle, episodic-to-semantic consolidation, and dual-process theories (System 1-vs-System 2), revealing the brain as a homology engine that minimizes topological complexity to transmute high-entropy sensory flux into low-entropy, invariant cognitive structure.

Large language models (LLMs) have achieved state-of-the-art performance in a variety of tasks, but remain largely opaque in terms of their internal mechanisms. Understanding these mechanisms is crucial to improve their reasoning abilities. Drawing inspiration from the interplay between neural processes and human cognition, we propose a novel interpretability framework to systematically analyze the roles and behaviors of attention heads, which are key components of LLMs. We introduce CogQA, a dataset that decomposes complex questions into step-by-step subquestions with a chain-of-thought design, each associated with specific cognitive functions such as retrieval or logical reasoning. By applying a multi-class probing method, we identify the attention heads responsible for these functions. Our analysis across multiple LLM families reveals that attention heads exhibit functional specialization, characterized as cognitive heads. These cognitive heads exhibit several key properties: they are universally sparse, vary in number and distribution across different cognitive functions, and display interactive and hierarchical structures. We further show that cognitive heads play a vital role in reasoning tasks - removing them leads to performance degradation, while augmenting them enhances reasoning accuracy. These insights offer a deeper understanding of LLM reasoning and suggest important implications for model design, training, and fine-tuning strategies.

Alzheimer's disease (AD) persists as a paramount challenge in neurological research, characterized by the pathological hallmarks of amyloid-beta (Abeta) plaques and neurofibrillary tangles composed of hyperphosphorylated tau. This review synthesizes the evolving understanding of AD pathogenesis, moving beyond the linear amyloid cascade hypothesis to conceptualize the disease as a cross-talk of intricately interacting pathologies, encompassing Abeta, tau, and neuroinflammation. This evolving pathophysiological understanding parallels a transformation in diagnostic paradigms, where biomarker-based strategies -- such as the AT(N) framework -- enable early disease detection during preclinical or prodromal stages. Within this new landscape, while anti-Abeta monoclonal antibodies (e.g., lecanemab, donanemab) represent a breakthrough as the first disease-modifying therapies, their modest efficacy underscores the limitation of single-target approaches. Therefore, this review explores the compelling rationale for combination therapies that simultaneously target Abeta pathology, aberrant tau, and neuroinflammation. Looking forward, we emphasize emerging technological platforms -- such as gene editing and biophysical neuromodulation -- n advancing precision medicine. Ultimately, the integration of early biomarker detection, multi-target therapeutic strategies, and AI-driven patient stratification charts a promising roadmap toward fundamentally altering the trajectory of AD. The future of AD management will be defined by preemptive, biomarker-guided, and personalized combination interventions.

Brain dynamics dominate every level of neural organization -- from single-neuron spiking to the macroscopic waves captured by fMRI, MEG, and EEG -- yet the mathematical tools used to interrogate those dynamics remain scattered across a patchwork of traditions. Neural mass models (NMMs) (aggregate neural models) provide one of the most popular gateways into this landscape, but their sheer variety -- spanning lumped parameter models, firing-rate equations, and multi-layer generators -- demands a unifying framework that situates diverse architectures along a continuum of abstraction and biological detail. Here, we start from the idea that oscillations originate from a simple push-pull interaction between two or more neural populations. We build from the undamped harmonic oscillator and, guided by a simple push-pull motif between excitatory and inhibitory populations, climb a systematic ladder of detail. Each rung is presented first in isolation, next under forcing, and then within a coupled network, reflecting the progression from single-node to whole-brain modeling. By transforming a repertoire of disparate formalisms into a navigable ladder, we hope to turn NMM choice from a subjective act into a principled design decision, helping both theorists and experimentalists translate between scales, modalities, and interventions. In doing so, we offer a \emph{Rosetta Stone} for brain oscillation models -- one that lets the field speak a common dynamical language while preserving the dialectical richness that fuels discovery.

Serotonin (5-hydroxytryptamine) is a major neurotransmitter whose release from densely distributed serotonergic varicosities shapes plasticity and network integration throughout the brain, yet its extracellular dynamics remain poorly understood due to the sub-micrometer and millisecond scales involved. We develop a mathematical framework that captures the coupled reaction-diffusion processes governing serotonin signaling in realistic tissue microenvironments. Formulating a two-dimensional compartmental-reaction diffusion system, we use strong localized perturbation theory to derive an asymptotically equivalent set of nonlinear integro-ODEs that preserve diffusive coupling while enabling efficient computation. We analyze period-averaged steady states, establish bounds using Jensen's inequality, obtain closed-form spike maxima and minima, and implement a fast marching-scheme solver based on sum-of-exponentials kernels. These mathematical results provide quantitative insight into how firing frequency, varicosity geometry, and uptake kinetics shape extracellular serotonin. The model reveals that varicosities form diffusively coupled microdomains capable of generating spatial "serotonin reservoirs," clarifies aspects of local versus volume transmission, and yields predictions relevant to interpreting high-resolution serotonin imaging and the actions of selective serotonin-reuptake inhibitors.

We propose Symmetry-Loss, a brain-inspired algorithmic principle that enforces invariance and equivariance through a differentiable constraint derived from environmental symmetries. The framework models learning as the iterative refinement of an effective symmetry group, paralleling developmental processes in which cortical representations align with the world's structure. By minimizing structural surprise, i.e. deviations from symmetry consistency, Symmetry-Loss operationalizes a Free-Energy--like objective for representation learning. This formulation bridges predictive-coding and group-theoretic perspectives, showing how efficient, stable, and compositional representations can emerge from symmetry-based self-organization. The result is a general computational mechanism linking developmental learning in the brain with principled representation learning in artificial systems.

The existence of 'what' and 'where' pathways of information processing in the brain was proposed almost 30 years ago, but there is still a lack of a clear mathematical model that could show how these pathways work together. We propose a biologically inspired mathematical model that uses this idea to identify and separate the self from the environment and then build and use a self-model for better predictions. This is a model of neocortical columns governed by the basal ganglia to make predictions and choose the next action, where some columns act as 'what' columns and others act as 'where' columns. Based on this model, we present a reinforcement learning agent that learns purposeful behavior in a virtual environment. We evaluate the agent on the Atari games Pong and Breakout, where it successfully learns to play. We conclude that the ability to separate the self from the environment gives advantages to the agent and therefore such a model could appear in living organisms during evolution. We propose Self-Awareness Principle 1: the ability to separate the self from the world is a necessary but insufficient condition for self-awareness.

In the process of evolution, the brain has achieved such perfection that artificial intelligence systems do not have and which needs its own mathematics. The concept of cognitome, introduced by the academician K.V. Anokhin, as the cognitive structure of the mind -- a high-order structure of the brain and a neural hypernetwork, is considered as the basis for modeling. Consciousness then is a special form of dynamics in this hypernetwork -- a large-scale integration of its cognitive elements. The cognitome, in turn, consists of interconnected COGs (cognitive groups of neurons) of two types -- functional systems and cellular ensembles. K.V. Anokhin sees the task of the fundamental theory of the brain and mind in describing these structures, their origin, functions and processes in them. The paper presents mathematical models of these structures based on new mathematical results, as well as models of different cognitive processes in terms of these models. In addition, it is shown that these models can be derived based on a fairly general principle of the brain works: \textit{the brain discovers all possible causal relationships in the external world and draws all possible conclusions from them}. Based on these results, the paper presents models of: ``natural" classification; theory of functional brain systems by P.K. Anokhin; prototypical theory of categorization by E. Roche; theory of causal models by Bob Rehter; theory of consciousness as integrated information by G. Tononi.

Cancer patients may undergo lengthy and painful chemotherapy treatments, comprising several successive regimens or plans. Treatment inefficacy and other adverse events can lead to discontinuation (or failure) of these plans, or prematurely changing them, which results in a significant amount of physical, financial, and emotional toxicity to the patients and their families. In this work, we build treatment failure models based on the Real World Evidence (RWE) gathered from patients' profiles available in our oncology EMR/EHR system. We also describe our feature engineering pipeline, experimental methods, and valuable insights obtained about treatment failures from trained models. We report our findings on five primary cancer types with the most frequent treatment failures (or discontinuations) to build unique and novel feature vectors from the clinical notes, diagnoses, and medications that are available in our oncology EMR. After following a novel three axes - performance, complexity and explainability - design exploration framework, boosted random forests are selected because they provide a baseline accuracy of 80% and an F1 score of 75%, with reduced model complexity, thus making them more interpretable to and usable by oncologists.

Representations pervade our daily experience, from letters representing sounds to bit strings encoding digital files. While such representations require externally defined decoders to convey meaning, conscious experience appears fundamentally different: a neural state corresponding to perceiving a red square cannot alternatively encode the experience of a green square. This intrinsic property of consciousness suggests that conscious representations must be unambiguous in a way that conventional representations are not. We formalize this intuition using information theory, defining representational ambiguity as the conditional entropy H(I|R) over possible interpretations I given a representation R. Through experiments on neural networks trained to classify MNIST digits, we demonstrate that relational structures in network connectivity can unambiguously encode representational content. Using both learned decoders and direct geometric matching, we achieve perfect (100%) accuracy for dropout-trained networks and 38% for standard backpropagation in identifying output neuron class identity, despite identical task performance, demonstrating that representational ambiguity can arise orthogonally to behavioral accuracy. We further show that spatial position information of input neurons can be decoded from network connectivity with R2 up to 0.844. These results provide a quantitative method for measuring representational ambiguity in neural systems and demonstrate that neural networks can exhibit the low-ambiguity representations posited as necessary (though not sufficient) by theoretical accounts of consciousness.

Transmission dynamics of infectious diseases are often studied using compartmental mathematical models, which are commonly represented as systems of autonomous ordinary differential equations. A key step in the analysis of such models is to identify equilibria and find conditions for their stability. Local stability analysis reduces to a problem in linear algebra, but there is no general algorithm for establishing global stability properties. Substantial progress on global stability of epidemic models has been made in the last 20 years, primarily by successfully applying Lyapunov's method to specific systems. Here, we show that any compartmental epidemic model in which susceptible individuals cannot be distinguished and can be infected only once, has a globally asymptotically stable (GAS) equilibrium. If the basic reproduction number ${R}_0$ satisfies ${R}_0 > 1$, then the GAS fixed point is an endemic equilibrium (i.e., constant, positive disease prevalence). Alternatively, if ${R}_0 \le 1$, then the GAS equilibrium is disease-free. This theorem subsumes a large number of results published over the last century, strengthens most of them by establishing global rather than local stability, avoids the need for any stability analyses of these systems in the future, and settles the question of whether co-existing stable solutions or non-equilibrium attractors are possible in such models: they are not.

Honey bees play a crucial role in pollination, contributing significantly to global agriculture and ecosystems. Accurately estimating hive populations is essential for understanding the effects of environmental factors on bee colonies, yet traditional methods of counting bees are time-consuming, labor-intensive, and prone to human error, particularly in large-scale studies. In this paper, we present a deep learning-based solution for automating bee population counting using CSRNet and introduce ASUBEE, the FIRST high-resolution dataset specifically designed for this task. Our method employs density map estimation to predict bee populations, effectively addressing challenges such as occlusion and overlapping bees that are common in hive monitoring. We demonstrate that CSRNet achieves superior performance in terms of time efficiency, with a computation time of just 1 second per image, while delivering accurate counts even in complex and densely populated hive scenarios. Our findings show that deep learning approaches like CSRNet can dramatically enhance the efficiency of hive population assessments, providing a valuable tool for researchers and beekeepers alike. This work marks a significant advancement in applying AI technologies to ecological research, offering scalable and precise monitoring solutions for honey bee populations.

Agave villalobosii sp. nov. (sect. Ditepalae, Agavaceae, Asparagales) from the Mexican central plain in Aguascalientes and southern Zacatecas, Mexico, is described and illustrated. It resembles A. flexispina in terms of color and the general appearance of its rosettes. However, it differs from the latter in having fewer leaves with more widely spaced teeth, more compact and shorter panicles with more inclined lateral branches with respect to the horizontal plane, and subglobose to broadly ellipsoid capsules. A distribution map for both species is provided. The species was preliminary assessed as critically endangered.

Life demonstrates remarkable homochirality of its major building blocks: nucleic acids, amino acids, sugars, and phospholipids. We propose a new model that places the root of the life homochirality in the formation of protocellular bilayer vesicles (liposomes). These liposomes are formed at the water/air interface from Langmuir layers and contain ribose, which is known to be delivered by carbonaceous meteorites; hence, the model suggests that impact craters are likely loci of life origin. The ribose delivered by meteorites appeared as racemic, yet life is based on D-ribose and its derivatives. The high membrane permeability to D-ribose implies a strong interaction with the charged phosphate head groups of the bilayer membrane. Such interaction, along with the presence of Fe3+, leads to ribose phosphorylation, forming ribose-5-phosphate, a molecule that cannot cross the membrane. As a result, the D-ribose-5-phosphate formed inside the vesicle becomes trapped and cannot escape. Over time, the phosphorylated D-ribose accumulates inside vesicles, forming a population of D-ribose-5-phosphate vesicles. Following the selective accumulation of phosphorylated D-ribose, D-oxyribose-5-phosphate forms, facilitating the transition of entrapped D-ribose-5-phosphate molecules into complex functional molecules, such as ribozymes/RNA, and over time, into DNA. The proposed model can be tested experimentally.

Evolutionary graph theory (EGT) studies the effect of population structure on evolutionary dynamics. The vertices of the graph represent the $N$ individuals. The edges denote interactions for competitive replacement. Two standard update rules are death-Birth (dB) and Birth-death (Bd). Under dB, an individual is chosen uniformly at random to die, and its neighbors compete to fill the vacancy proportional to their fitness. Under Bd, an individual is chosen for reproduction proportional to fitness, and its offspring replaces a randomly chosen neighbor. Here we study mixed updating between those two scenarios. In each time step, with probability $\delta$ the update is dB and with remaining probability it is Bd. We study fixation probabilities and times as functions of $\delta$ under constant selection. Despite the fact that fixation probabilities and times can be increasing, decreasing, or non-monotonic in $\delta$, we prove nearly all unweighted undirected graphs have short fixation times and provide an efficient algorithm to estimate their fixation probabilities. We also prove that weighted directed graphs that are uniform circulations have fixation probability $1/N$ for every $\delta$. Finally, we prove exact formulas for fixation probabilities on cycles, stars, and more complex structures and classify their sensitivities to $\delta$.

Endangered populations often experience limited growth ability at low densities, a phenomenon described by the Allee effect. In this thesis, we investigate a predator-prey model incorporating the Allee effect within a two-dimensional nonlinear reaction-diffusion framework, with the aim of understanding how local spatial refuges can promote the persistence of low-density populations by enabling them to surpass recovery thresholds. We first simulate an extinction-prone scenario in which initial densities fall below the Allee threshold, demonstrating that most populations tend toward extinction. We then introduce protected areas together with positive growth terms to facilitate survival. To assess the role of diffusion-reaction dynamics, we construct an objective function based on the shape and location of protected areas, and employ a bi-objective optimization approach. Our results reveal that as the weights of the objective function vary, the optimal protectedarea configuration shifts between fragmented and contiguous patterns. We begin with a single-species prey analysis and subsequently extend the model to include predators, where we use mathematical analysis to investigate the steady states of the two-species system.

HIV pre-exposure prophylaxis (PrEP) drastically reduces the risk of HIV infection if taken as prescribed, providing almost perfect protection even during unprotected sexual intercourse. Although this has been transformative in reducing new HIV infections among high-risk populations, it has also been linked to an increase in risk practices -- a phenomenon known as risk compensation -- thereby favoring the spread of other sexually transmitted infections (STIs) deemed less severe. In this paper, we study a minimal compartmental model describing the effect of risk awareness and risk compensation due to PrEP on the spread of other STIs among a high-infection-risk group of men who have sex with men (MSM). The model integrates three key elements of risk-mediated behavior and PrEP programs: (i) HIV risk awareness drives self-protective behaviors (such as condom use and voluntary STI screening); (ii) individuals on PrEP are subject to risk compensation, but (iii) are required to screen for asymptomatic STIs frequently. We derived the basic reproduction number of the system, $R_0$, and found a transcritical bifurcation at $R_0=1$, where the disease-free equilibrium becomes unstable and an endemic equilibrium emerges. This endemic equilibrium is asymptotically stable wherever it exists. We identified critical thresholds in behavioral and policy parameters that separate these regimes and analyzed typical values for plausible parameter choices. Beyond the specific epidemiological context, the model serves as a general framework for studying nonlinear interactions between behavioral adaptation, preventive interventions, and disease dynamics, providing insights into how feedback mechanisms can lead to non-trivial responses in epidemic systems. Finally, our model can be easily extended to study the effect of interventions and risk compensation in other STIs.

Mitigation measures are essential for controlling the spread of infectious diseases during pandemics and epidemics, but they impose considerable societal, individual, and economic costs. We developed a general optimization framework to balance costs related to infection and to mitigation. Optimizing the trade-off between mitigation and infection cost, we identified three novel, surprising effects: First, assuming a constant reproduction number $R_0$, the optimal response to an infectious disease requires either strict mitigation or none at all, depending on disease severity, but never does one find an intermediate mitigation level to be optimal. Second, under seasonal variations, optimal mitigation is stricter during winter. Interestingly, a single wave of infections still arises in spring with 3 months delay to the seasonal peak of infectivity, replacing the autumn/winter waves known for classical influenza. Third, during steady vaccination campaigns, even optimal mitigation can result in transient infection waves. Finally, we quantify the cost of delayed mitigation onset and show that even short delays can substantially increase total costs -- if the disease is severe. Overall, our framework is easily applicable to general and complex settings and thereby presents a versatile tool to explore optimal mitigation strategies for endemic and pandemic infectious disease.

Protein S (PS) is a notable anticoagulant implicated in both bleeding and thrombotic disorders, making it a promising drug target. Importantly, PS enhances the anticoagulant function of TFPI$\alpha$, likely circulating in the bloodstream together with TFPI$\alpha$ and a truncated form of factor V (fVshort) in the trimolecular complex, TFPI$\alpha$-fVshort-PS, which we call protein S complex (PSC). PSC has been proposed to strongly inhibit thrombin production by enhancing the ability of TFPI$\alpha$ to inhibit clotting factor Xa up to 100-fold and by localizing to platelet membranes, limiting fXa activity shortly after coagulation starts. Yet, exactly how PS functions with TFPI$\alpha$ as an anticoagulant remains poorly understood. To investigate, we extend an experimentally validated mathematical model of blood coagulation to include PSC and free PS (not part of PSC) in the plasma, as well as free PS and TFPI$\alpha$ in platelets. We find that shortly after coagulation initiation, PSC strongly inhibits thrombin production. We find that the (unknown) magnitude of the enhanced affinity of PSC binding to inhibit fXa critically regulates PSC's impact on thrombin production. We find that under flow, PSC can unexpectedly accumulate on platelets to concentrations ~50 times higher than in the plasma. We also find that PSC limits thrombin production by occupying fV-specific binding sites on platelets. Our results show that changes in PSC can dramatically impact severity of pathological bleeding disorders. For the east Texas bleeding disorder, elevated PSC concentrations eliminate thrombin bursts, leading to bleeding. With fV deficiency, reducing PSC rescues thrombin production in severe fV deficiency and returns thrombin production due to mild fV deficiency to normal. Finally, thrombin production in severe hemophilia A can be substantially improved by blocking PSC's anticoagulant function.

Introduction: The blood-brain barrier (BBB) protects the central nervous system but prevents most neurotherapeutics from reaching effective concentrations in the brain. BBB-penetrating peptides (BBBPs) offer a promising strategy for brain drug delivery; however, the scarcity of positive samples and severe class imbalance hinder the reliable identification of BBBPs. Objectives: Our goal is to alleviate class imbalance in BBBP prediction and to develop an accurate, interpretable classifier for BBBP prediction. Methods: We propose DREAM-B3P, which couples a feedback diffusion model (FB-Diffusion) for data augmentation with a dual-stream Transformer for classification. FB-Diffusion learns the BBBP distribution via iterative denoising and uses an external analyzer to provide feedback, generating high-quality pseudo-BBBPs. The classifier contains a sequence stream that extracts structural features from peptide sequences and a physicochemical stream that captures physicochemical features such as hydrophobic surface area, molecular charge, number of rotatable bonds, and polarizability. Combining the two features leads to superior BBBP predictive performance. Results: On a benchmark test set containing equal numbers of BBBPs and non-BBBPs, DREAM-B3P surpasses baseline methods (Deep-B3P, B3Pred, BBPpredict and Augur), improving AUC/ACC/MCC by 4.3\%/17.8\%/14.9\%, respectively, over the second-best method. Conclusion: By integrating feedback diffusion with a dual-stream Transformer classifier, DREAM-B3P effectively mitigates data scarcity and imbalance and achieves state-of-the-art performance.

Rapid advancements in technology have led to an increased use of artificial intelligence (AI) technologies in medicine and bioinformatics research. In anticipation of this, the National Institutes of Health (NIH) assembled the Bridge to Artificial Intelligence (Bridge2AI) consortium to coordinate development of AI-ready datasets that can be leveraged by AI models to address grand challenges in human health and disease. The widespread availability of genome sequencing technologies for biomedical research presents a key data type for informing AI models, necessitating that genomics data sets are AI-ready. To this end, the Genomic Information Standards Team (GIST) of the Bridge2AI Standards Working Group has documented a set of recommendations for maintaining AI-ready genomics datasets. In this report, we describe recommendations for the collection, storage, identification, and proper use of genomics datasets to enable them to be considered AI-ready and thus drive new insights in medicine through AI and machine learning applications.

Precise irrigation management requires robust classification of plant water stress. We expanded a morpho-kinematic (MK) framework that derives canopy-movement features from RGB time-lapse imaging evaluating how methodological refinements affect robustness and fine discrimination across four irrigation treatments representing distinct stress histories. The study tested both a biological (Agg) versus an isogonal (Unif) sectoring of the canopy image, within an additive scheme where to the baseline (i.e. flattened MK features, A0) were sequentially added non-linear descriptors (A1), irrigation-context variables (i.e. dry time, A2), and their interactions with baseline (A3). The multi-class problem was decomposed in biologically meaningful binary tasks, and the final classification confronted an adaptive - to the performance obtained in the out-of-fold predictions inside the leave-one-sample-out validation framework - linear opinion pooling (ALOP) ensemble, evaluated across its full parameter space, against hierarchical cascades (HCC). In our combined dataset from two sequential Lactuca sativa experiments (144 sample-days) ALOP median outperformed HCC in every configuration, while non-linear and contextual enrichments (A1-A2) produced consistent improvements in terms of prediction stability, variability (for ALOP), and balanced accuracy (BA). The highest balanced accuracy (median BA = 0.96) was reached under Unif scheme in A3, yet the Agg configuration in A2 achieved the best compromise between accuracy (median BA approx 0.91) and robustness. Concluding, this study identifies methodological pathways that strengthen resilience and transferability of movement-based water-stress classification, establishing a solid foundation for generalizable, low-cost phenotyping.

We introduce a method for approximating posterior probabilities of phylogenetic trees and reconstructing ancestral sequences under models of sequence evolution with site-dependence, where standard phylogenetic likelihood computations (pruning) fail. Our approach uses a combined data-augmentation and importance sampling scheme. A key advantage of our approach is the ability to leverage existing highly optimized phylogenetic software. We apply our approach to the reconstruction of B cell receptor affinity maturation lineages from high-throughput repertoire sequencing data and evaluate the impact of incorporating site-dependence on the reconstruction accuracy of both trees and ancestral sequences. We show that accounting for context-dependence during inference always improves the estimates of both ancestral sequences and lineage trees on simulated datasets. We also examine the impact of incorporating priors based on VDJ recombination models, and find that they significantly improve ancestral sequence reconstruction in germline-encoded regions, but increase errors in non-templated nucleotides. We propose a modified, piecewise prior to address this demonstrate that it improves empirical reconstruction accuracy. We apply our approach to the analysis of the HIV broadly neutralizing antibodies DH270 and CH235 which are important targets of current vaccine design efforts.

Multimodal MRI offers complementary multi-scale information to characterize the brain structure. However, it remains challenging to effectively integrate multimodal MRI while achieving neuroscience interpretability. Here we propose to use Laplacian harmonics and spectral graph theory for multimodal alignment and multiscale integration. Based on the cortical mesh and connectome matrix that offer multi-scale representations, we devise Laplacian operators and spectral graph attentions to construct a shared latent space for model alignment. Next, we employ a disentangled learning combined with Graph Variational Autoencoder architectures to separate scale-specific and shared features. Lastly, we design a mutual information-informed bilevel regularizer to separate causal and non-causal factors based on the disentangled features, achieving robust model performance with enhanced interpretability. Our model outperforms baselines and other state-of-the-art models. The ablation studies confirmed the effectiveness of the proposed modules. Our model promises to offer a robust and interpretable framework for multi-scale brain structure analysis.

Generating precise 3D molecular geometries is crucial for drug discovery and material science. While prior efforts leverage 1D representations like SELFIES to ensure molecular validity, they fail to fully exploit the rich chemical knowledge entangled within 1D models, leading to a disconnect between 1D syntactic generation and 3D geometric realization. To bridge this gap, we propose MolSculpt, a novel framework that "sculpts" 3D molecular geometries from chemical syntax. MolSculpt is built upon a frozen 1D molecular foundation model and a 3D molecular diffusion model. We introduce a set of learnable queries to extract inherent chemical knowledge from the foundation model, and a trainable projector then injects this cross-modal information into the conditioning space of the diffusion model to guide the 3D geometry generation. In this way, our model deeply integrates 1D latent chemical knowledge into the 3D generation process through end-to-end optimization. Experiments demonstrate that MolSculpt achieves state-of-the-art (SOTA) performance in \textit{de novo} 3D molecule generation and conditional 3D molecule generation, showing superior 3D fidelity and stability on both the GEOM-DRUGS and QM9 datasets. Code is available at this https URL.

We investigate multicellular sender receiver systems embedded in hydrogel beads, where diffusible signals mediate interactions among heterogeneous cells. Such systems are modeled by PDE ODE couplings that combine three dimensional diffusion with nonlinear intracellular dynamics, making analysis and simulation challenging. We show that the diffusion dynamics converges exponentially to a quasi steady spatial profile and use singular perturbation theory to reduce the model to a finite dimensional multiagent network. A closed form communication matrix derived from the spherical Green's function captures the effective sender receiver coupling. Numerical results show the reduced model closely matches the full dynamics while enabling scalable simulation of large cell populations.

This paper provides global attractivity results for the interior equilibrium point of a general Lotka-Volterra system with no restriction on the dimension of the system and with no special structure or properties of the interaction matrix. The main result contains as special cases all known general results, including the Volterra-Lyapunov theorem and the recently proposed eigenvector conditions. Moreover, global attractivity of the interior equilibrium point is shown for a three-dimensional example, where none of the existing general results can be applied.

Reliable in silico molecular toxicity prediction is a cornerstone of modern drug discovery, offering a scalable alternative to experimental screening. However, the black-box nature of state-of-the-art models remains a significant barrier to adoption, as high-stakes safety decisions demand verifiable structural insights alongside predictive performance. To address this, we propose a novel multi-task learning (MTL) framework designed to jointly enhance accuracy and interpretability. Our architecture integrates a shared chemical language model with task-specific attention modules. By imposing an L1 sparsity penalty on these modules, the framework is constrained to focus on a minimal set of salient molecular fragments for each distinct toxicity endpoint. The resulting framework is trained end-to-end and is readily adaptable to various transformer-based backbones. Evaluated on the ClinTox, SIDER, and Tox21 benchmark datasets, our approach consistently outperforms both single-task and standard MTL baselines. Crucially, the sparse attention weights provide chemically intuitive visualizations that reveal the specific fragments influencing predictions, thereby enhancing insight into the model's decision-making process.

The development of foundation models for functional magnetic resonance imaging (fMRI) time series holds significant promise for predicting phenotypes related to disease and cognition. Current models, however, are often trained using a mask-and-reconstruct objective on small brain regions. This focus on low-level information leads to representations that are sensitive to noise and temporal fluctuations, necessitating extensive fine-tuning for downstream tasks. We introduce Brain-Semantoks, a self-supervised framework designed specifically to learn abstract representations of brain dynamics. Its architecture is built on two core innovations: a semantic tokenizer that aggregates noisy regional signals into robust tokens representing functional networks, and a self-distillation objective that enforces representational stability across time. We show that this objective is stabilized through a novel training curriculum, ensuring the model robustly learns meaningful features from low signal-to-noise time series. We demonstrate that learned representations enable strong performance on a variety of downstream tasks even when only using a linear probe. Furthermore, we provide comprehensive scaling analyses indicating more unlabeled data reliably results in out-of-distribution performance gains without domain adaptation.

Innovation in biomaterials has brought both breakthroughs and new challenges in medicine, as implant materials have become increasingly multifunctional and complex. One of the greatest issues is the difficulty in assessing the temporal and multidimensional dynamics of tissue-implant interactions. Implant biology remains hard to decipher without a noninvasive and multiplexed technique that can accurately monitor real-time biological processes. To address this, we developed a multifunctional, self-sensing implant material composed of gold nano-columns patterned on a titanium surface (AuNC-Ti). This material acts as a nanoengineered surface-enhanced Raman spectroscopy (SERS) substrate that amplifies biological Raman signals at the tissue-implant interface, providing the ability to sense tissue-material interactions in a multiplexed and nondestructive manner. AuNC-Ti SERS substrates were fabricated using oblique angle deposition (OAD) and characterized using scanning electron microscopy (SEM) to show uniform formation of AuNCs ($360 \pm 40$ nm in length and $50 \pm 16$ nm in width). X-ray photoelectron spectroscopy (XPS), X-ray diffraction (XRD), and contact angle measurements demonstrated biocompatible surface chemistry with ideal wettability. Biocompatibility was further demonstrated via in vitro cytotoxicity assays on human aortic endothelial cells (HAECs) cultured on AuNC-Ti surfaces. The median SERS enhancement factor (EF) was calculated to be $1.8 \times 10^5$, and spatial identification of reporter molecules and porcine tissue components on AuNC-Ti surfaces was demonstrated using confocal Raman imaging and multivariate analysis. Our approach utilizes unlabeled SERS and machine learning, promising multiplexed characterization of tissue-material interactions and subsequently enabling tissue state determination and non-invasive monitoring of implant-tissue interaction.

The delimitation of biological species, i.e., deciding which individuals belong to the same species and whether and how many different species are represented in a data set, is key to the conservation of biodiversity. Much existing work uses only genetic data for species delimitation, often employing some kind of cluster analysis. This can be misleading, because geographically distant groups of individuals can be genetically quite different even if they belong to the same species. We investigate the problem of testing whether two potentially separated groups of individuals can belong to a single species or not based on genetic and spatial data. Existing methods such as the partial Mantel test and jackknife-based distance-distance regression are considered. New approaches, i.e., an adaptation of a mixed effects model, a bootstrap approach, and a jackknife version of partial Mantel, are proposed. All these methods address the issue that distance data violate the independence assumption for standard inference regarding correlation and regression; a standard linear regression is also considered. The approaches are compared on simulated meta-populations generated with SLiM and GSpace - two software packages that can simulate spatially-explicit genetic data at an individual level. Simulations show that the new jackknife version of the partial Mantel test provides a good compromise between power and respecting the nominal type I error rate. Mixed-effects models have larger power than jackknife-based methods, but tend to display type I error rates slightly above the significance level. An application on brassy ringlets concludes the paper.

Single-cell transcriptomics and proteomics have become a great source for data-driven insights into biology, enabling the use of advanced deep learning methods to understand cellular heterogeneity and gene expression at the single-cell level. With the advent of spatial-omics data, we have the promise of characterizing cells within their tissue context as it provides both spatial coordinates and intra-cellular transcriptional or protein counts. Proteomics offers a complementary view by directly measuring proteins, which are the primary effectors of cellular function and key therapeutic targets. However, existing models either ignore the spatial information or the complex genetic and proteomic programs within cells. Thus they cannot infer how cell internal regulation adapts to microenvironmental cues. Furthermore, these models often utilize fixed gene vocabularies, hindering their generalizability unseen genes. In this paper, we introduce HEIST, a hierarchical graph transformer foundation model for spatial transcriptomics and proteomics. HEIST models tissues as hierarchical graphs. The higher level graph is a spatial cell graph, and each cell in turn, is represented by its lower level gene co-expression network graph. HEIST achieves this by performing both intra-level and cross-level message passing to utilize the hierarchy in its embeddings and can thus generalize to novel datatypes including spatial proteomics without retraining. HEIST is pretrained on 22.3M cells from 124 tissues across 15 organs using spatially-aware contrastive and masked autoencoding objectives. Unsupervised analysis of HEIST embeddings reveals spatially informed subpopulations missed by prior models. Downstream evaluations demonstrate generalizability to proteomics data and state-of-the-art performance in clinical outcome prediction, cell type annotation, and gene imputation across multiple technologies.

Our adaptive immune system relies on the persistence over long times of a diverse set of antigen-experienced B cells to encode our memories of past infections and to protect us against future ones. While longitudinal repertoire sequencing promises to track the long-term dynamics of many B cell clones simultaneously, sampling and experimental noise make it hard to draw reliable quantitative conclusions. Leveraging statistical inference, we infer the dynamics of memory B cell clonal dynamics and conversion to plasmablasts, which includes clone creation, degradation, abundance fluctuations, and differentiation. We find that memory B cell clones degrade slowly, with a half-life of 10 years. Based on the inferred parameters, we predict that it takes about 50 years to renew 50\% of the repertoire, with most observed clones surviving for a lifetime. We infer that, on average, 1 out of 100 memory B cells differentiates into a plasmablast each year, more than expected from purely antigen-stimulated differentiation, and that plasmablast clones degrade with a half-life of about one year in the absence of memory imports. Our method is general and could be applied to other longitudinal repertoire sequencing B cell subsets.

Global pandemics, such as the recent COVID-19 crisis, highlight the need for stochastic epidemic models that can capture the randomness inherent in the spread of disease. Such models must be accompanied by methods for estimating parameters in order to generate fast nowcasts and short-term forecasts that can inform public health decisions. This paper presents a comparison of two advanced Bayesian inference methods: 1) pseudo-marginal particle Markov chain Monte Carlo, short Particle Filters (PF), and 2) Conditional Normalizing Flows (CNF). We investigate their performance on two commonly used compartmental models: a classical Susceptible-Infected-Recovered (SIR) model and a two-variant Susceptible-Exposed-Infected-Recovered (SEIR) model, complemented by an observation model that maps latent trajectories to empirical data. Addressing the challenges of intractable likelihoods for parameter inference in stochastic settings, our analysis highlights how these likelihood-free methods provide accurate and robust inference capabilities. The results of our simulation study further underscore the effectiveness of these approaches in capturing the stochastic dynamics of epidemics, providing prediction capabilities for the control of epidemic outbreaks. Results on an Ethiopian cohort study demonstrate operational robustness under real-world noise and irregular data sampling. To facilitate reuse and to enable building pipelines that ultimately contribute to better informed decision making in public health, we make code and synthetic datasets publicly available.

We present a computational framework that integrates functional-structural plant modeling (FSPM) with an evolutionary algorithm to optimize three-dimensional maize canopy architecture for enhanced light interception under high-density planting. The optimization revealed an emergent ideotype characterized by two distinct strategies: a vertically stratified leaf profile (steep, narrow upper leaves for penetration; broad, horizontal lower leaves for capture) and a radially tiled azimuthal arrangement that breaks the conventional distichous symmetry of maize to minimize self and mutual shading. Reverse ray-tracing simulations show that this architecture intercepts significantly more photosynthetically active radiation (PAR) than virtual canopies parameterized from high-performing field hybrids, with gains that generalize across multiple U.S. latitudes and planting densities. The optimized trait combinations align with characteristics of modern density-tolerant cultivars, supporting biological plausibility. Because recent gene editing advances enable more independent control of architectural traits, the designs identified here are increasingly feasible. By uncovering effective, non-intuitive trait configurations, our approach provides a scalable, predictive tool to guide breeding targets, improve light-use efficiency, and ultimately support sustainable yield gains.

We are interested in prey-predator communities where the predator population evolves much faster than the prey's (e.g. insect-tree communities). We introduce a piecewise deterministic model for these prey-predator communities that arises as a limit of a microscopic model when the number of predators goes to infinity. We prove that the process has a unique invariant probability measure and that it is exponentially ergodic. Further on, we rescale the predator dynamics in order to model predators of smaller size. This slow-fast system converges to a community process in which the prey dynamics is averaged on the predator equilibria. This averaged process admits an invariant probability measure which can be computed explicitly. We use numerical simulations to study the convergence of the invariant probability measures of the rescaled processes.

Mathematical modeling in systems toxicology enables a comprehensive understanding of the effects of pharmaceutical substances on cardiac health. However, the complexity of these models limits their widespread application in early drug discovery. In this paper, we introduce a novel approach to solving parameterized models of cardiac action potentials by combining meta-learning techniques with Systems Biology-Informed Neural Networks (SBINNs). The proposed method, hyperSBINN, effectively addresses the challenge of predicting the effects of various compounds at different concentrations on cardiac action potentials, outperforming traditional differential equation solvers in speed. Our model efficiently handles scenarios with limited data and complex parameterized differential equations. The hyperSBINN model demonstrates robust performance in predicting APD90 values, indicating its potential as a reliable tool for modeling cardiac electrophysiology and aiding in preclinical drug development. This framework represents an advancement in computational modeling, offering a scalable and efficient solution for simulating and understanding complex biological systems.

Data integration methods aim to extract low-dimensional embeddings from high-dimensional outcomes to remove unwanted variations, such as batch effects and unmeasured covariates, across heterogeneous datasets. However, multiple hypothesis testing after integration can be biased due to data-dependent processes. We introduce a robust post-integrated inference method that accounts for latent heterogeneity by utilizing control outcomes. Leveraging causal interpretations, we derive nonparametric identifiability of the direct effects using negative control outcomes. By utilizing surrogate control outcomes as an extension of negative control outcomes, we develop semiparametric inference on projected direct effect estimands, accounting for hidden mediators, confounders, and moderators. These estimands remain statistically meaningful under model misspecifications and with error-prone embeddings. We provide bias quantifications and finite-sample linear expansions with uniform concentration bounds. The proposed doubly robust estimators are consistent and efficient under minimal assumptions and potential misspecification, facilitating data-adaptive estimation with machine learning algorithms. Our proposal is evaluated using random forests through simulations and analysis of single-cell CRISPR perturbed datasets, which may contain potential unmeasured confounders.

We study a population of $N$ individuals evolving according to a biparental Moran model with two types, one being advantaged compared to the other. The advantage is conferred by a Mendelian mutation, that reduces the death probability of individuals carrying it. We assume that a proportion $a$ of individuals initially carry this mutation, which therefore eventually gets fixed with high probability. After a long time, we sample a gene uniformly from the population, at a new locus, independent of the locus under selection, and calculate the probability that this gene originated from one of the initially advantaged individuals, when the population size is large. Our theorem provides quantitative insights, such as the observation that under strong viability selection, if only $1\%$ of the individuals are initially advantaged, up to $19\%$ of the population's genome will originate from them after a long time.

Proteins are the essential drivers of biological processes. At the molecular level, they are chains of amino acids that can be viewed through a linguistic lens where the twenty standard residues serve as an alphabet combining to form a complex language, referred to as the language of life. To understand this language, we must first identify its fundamental units. Analogous to words, these units are hypothesized to represent an intermediate layer between single residues and larger domains. Crucially, just as protein diversity arises from evolution, these units should inherently reflect evolutionary relationships. We introduce PUMA (Protein Units via Mutation-Aware Merging) to discover these evolutionarily meaningful units. PUMA employs an iterative merging algorithm guided by substitution matrices to identify protein units and organize them into families linked by plausible mutations. This process creates a hierarchical genealogy where parent units and their mutational variants coexist, simultaneously producing a unit vocabulary and the genealogical structure connecting them. We validate that PUMA families are biologically meaningful; mutations within a PUMA family correlate with clinically benign variants and with high-scoring mutations in high-throughput assays. Furthermore, these units align with the contextual preferences of protein language models and map to known functional annotations. PUMA's genealogical framework provides evolutionarily grounded units, offering a structured approach for understanding the language of life.

We present a modular framework powered by large language models (LLMs) that automates and streamlines key tasks across the early-stage computational drug discovery pipeline. By combining LLM reasoning with domain-specific tools, the framework performs biomedical data retrieval, literature-grounded question answering via retrieval-augmented generation, molecular generation, multi-property prediction, property-aware molecular refinement, and 3D protein-ligand structure generation. The agent autonomously retrieved relevant biomolecular information, including FASTA sequences, SMILES representations, and literature, and answered mechanistic questions with improved contextual accuracy compared to standard LLMs. It then generated chemically diverse seed molecules and predicted 75 properties, including ADMET-related and general physicochemical descriptors, which guided iterative molecular refinement. Across two refinement rounds, the number of molecules with QED > 0.6 increased from 34 to 55. The number of molecules satisfying empirical drug-likeness filters also rose; for example, compliance with the Ghose filter increased from 32 to 55 within a pool of 100 molecules. The framework also employed Boltz-2 to generate 3D protein-ligand complexes and provide rapid binding affinity estimates for candidate compounds. These results demonstrate that the approach effectively supports molecular screening, prioritization, and structure evaluation. Its modular design enables flexible integration of evolving tools and models, providing a scalable foundation for AI-assisted therapeutic discovery.

Bulk tissue RNA sequencing of heterogeneous samples provides averaged gene expression profiles, obscuring cell type-specific dynamics. To address this, we present a probabilistic hierarchical Bayesian model that deconvolves bulk RNA-seq data into constituent cell-type expression profiles and proportions, leveraging a high-resolution single-cell reference. We apply our model to human endometrial tissue across the menstrual cycle, a context characterized by dramatic hormone-driven cellular composition changes. Our extended framework provides a principled inference of cell type proportions and cell-specific gene expression changes across cycle phases. We demonstrate the model's structure, priors, and inference strategy in detail, and we validate its performance with simulations and comparisons to existing methods. The results reveal dynamic shifts in epithelial, stromal, and immune cell fractions between menstrual phases, and identify cell-type-specific differential gene expression associated with endometrial function (e.g., decidualization markers in stromal cells during the secretory phase). We further conduct robustness tests and show that our Bayesian approach is resilient to reference mismatches and noise. Finally, we discuss the biological significance of our findings, potential clinical implications for fertility and endometrial disorders, and future directions, including integration of spatial transcriptomics.

Accurate simulations of electric fields (E-fields) in brain stimulation depend on tissue conductivity representations that link macroscopic assumptions with underlying microscopic tissue structure. Mesoscale conductivity variations can produce meaningful changes in E-fields and neural activation thresholds but remain largely absent from standard macroscopic models. Recent microscopic models have suggested substantial local E-field perturbations and could, in principle, inform mesoscale conductivity. However, the quantitative validity of microscopic models is limited by fixation-related tissue distortion and incomplete extracellular-space reconstruction. We outline approaches that bridge macro- and microscales to derive consistent mesoscale conductivity distributions, providing a foundation for accurate multiscale models of E-fields and neural activation in brain stimulation.