We present Hecate, a modular lossless genomic compression framework. It is designed around uncommon but practical source-coding choices. Unlike many single-method compressors, Hecate treats compression as a conditional coding problem over coupled FASTA/FASTQ streams (control, headers, nucleotides, case, quality, extras). It uses per-stream codecs under a shared indexed block container. Codecs include alphabet-aware packing with an explicit side channel for out-of-alphabet residues, an auxiliary-index Burrows-Wheeler pipeline with custom arithmetic coding, and a blockwise Markov mixture coder with explicit model-competition signaling. This architecture yields high throughput, exact random-access slicing, and referential mode through streamwise binary differencing. In a comprehensive benchmark suite, Hecate provides the best compression vs. speed trade-offs against state-of-the-art established tools (MFCompress, NAF, bzip3, AGC), with notably stronger behaviour on large genomes and high-similarity referential settings. For the same compression ratio, Hecate is 2 to 10 times faster. When given the same time budget as other algorithms, Hecate achieves up to 5% to 10% better compression.

Neuroscience has long informed the development of artificial neural networks, but the success of modern architectures invites, in turn, the converse: can modern networks teach us lessons about brain function? Here, we examine the structure of the cortical column and propose that the transformer provides a natural computational analogy for multiple elements of cortical microcircuit organization. Rather than claiming a literal implementation of transformer equations in cortex, we develop a hypothetical mapping between transformer operations and laminar cortical features, using the analogy as an orienting framework for analysis and discussion. This mapping allows us to examine in greater depth how contextual selection, content routing, recurrent integration, and interlaminar transformations may be distributed across cortical circuitry. In doing so, we generate a broad set of predictions and experimentally testable hypotheses concerning laminar specialization, contextual modulation, dendritic integration, oscillatory coordination, and the effective connectivity of cortical columns. This proposal is intended as a structured hypothesis rather than a definitive account of cortical computation. Placing transformer operations and cortical architectonics into a common descriptive framework sharpens questions, reveals new functional correspondences, and opens a productive route for reciprocal exchange between systems neuroscience and modern AI. More broadly, this perspective suggests that comparing brains and architectures at the level of computational organization can yield genuine insight into both.

This study investigates the cognitive motor control detection and the underlying neuroregulatory mechanisms during music-assisted simulated driving. Using a dynamic higher-order network model constructed with EEG-based cross-information entropy, we quantify the dynamic coordination within brain networks activated during both music listening and driving. This approach, which contrasts with previous static network analyses, provides novel insights into how musical stimuli modulate the complex interplay of brain regions during demanding tasks. Results demonstrated enhanced third-order connectivity and elevated higher-order information entropy in music-stimulated driving compared to baseline driving, as evidenced by increasing Phi values of higher-order network indices. Supervised machine learning, including support vector machines, revealed a strong correlation between model accuracy and ROC-AUC values and the hierarchy of brain network features. This underscores the importance of higher-order features in decoding brain motor-control states during music-simulated driving. These findings deepen our understanding of the interplay between music cognition and motor control, offering valuable insights for the development of novel brain-computer-music interfaces (BCMI) and adaptive human-machine systems to enhance performance in demanding tasks like driving.

We develop a generally covariant description of evolutionary dynamics that operates consistently in both genotype and phenotype spaces. We show that the maximum entropy principle yields a fundamental identification between the inverse metric tensor and the covariance matrix, revealing the Lande equation as a covariant gradient ascent equation. This demonstrates that evolution can be modeled as a learning process on the fitness landscape, with the specific learning algorithm determined by the functional relation between the metric tensor and the noise covariance arising from microscopic dynamics. While the metric (or the inverse genotypic covariance matrix) has been extensively characterized empirically, the noise covariance and its associated observable (the covariance of evolutionary changes) have never been directly measured. This poses the experimental challenge of determining the functional form relating metric to noise covariance.

The processes of morphogenesis that give rise to the shapes of organs and organisms during development are often driven by mechanical instabilities. Can such mechanical bifurcations also drive or constrain the evolution of these processes in the first place? We discover an instance of these constraints in the green algae of the family Volvocaceae. During their development, their bowl-shaped embryonic cell sheet turns itself inside out. This inversion is driven by a simple wave of cell wedging in the genus Pleodorina (16-128 cells) and more complex programmes of cell shape changes in Volvox (~400-50000 cells). However, no species with intermediate cell numbers (256 cells) have been described. Here, we relate this gap to a mechanical bifurcation: Focusing on the inversion of Pleodorina californica (64 cells), we develop a continuum model, in which the cell shape changes driving inversion appear as changes of the intrinsic curvature of an elastic surface. A mechanical bifurcation in this model predicts that inversion is only possible in a subset of its parameter space. Strikingly, parameters estimated for P. californica fall into this possible subset, but those that we extrapolate to 256 or more cells using allometric observations and a model of cell cleavage in Volvocaceae do not. Our work thus suggests that the more complex inversion strategies of Volvox are an evolutionary necessity to obviate this bifurcation and indicates more broadly how mechanical bifurcations can drive the evolution of morphogenesis.

Microbial communities harbor extensive fine-scale diversity: closely-related strains of the same species coexist alongside many distantly-related taxa. Yet strain coexistence remains poorly understood, largely because most studies neglect the diverse communities in which strains are embedded. Here we combine community ecology and statistical physics to study the dynamics of closely-related strains in a community context. We demonstrate that in a diverse community, indirect interactions between strains -- mediated through the surrounding community members -- can be as strong as direct ones. These community-mediated feedbacks cause conspecific strains to behave as if they have correlated growth rates and reduced competition. Using modern coexistence theory, we show that these effects correspond to equalizing and stabilizing mechanisms which together promote strain coexistence. The same equalizing and stabilizing mechanisms also qualitatively transform strain abundance correlations: strains that compete strongly and show negative correlations in isolation instead show positive correlations in a community, appearing mutualistic despite being competitors. Our results demonstrate that strain dynamics are emergent consequences of the surrounding community, and that capturing community feedbacks does not require the full interaction network; only a small number of emergent parameters.

Motivation: The scalable identification of bioactive compounds is essential for contemporary drug discovery. This process faces a key trade-off: structural screening offers scalability but lacks biological context, whereas high-content phenotypic profiling provides deep biological insights but is resource-intensive. The primary challenge is to extract robust biological signals from noisy data and encode them into representations that do not require biological data at inference. Results: This study presents DECODE (DEcomposing Cellular Observations of Drug Effects), a framework that bridges this gap by empowering chemical representations with intrinsic biological semantics to enable structure-based in silico biological profiling. DECODE leverages limited paired transcriptomic and morphological data as supervisory signals during training, enabling the extraction of a measurement-invariant biological fingerprint from chemical structures and explicit filtering of experimental noise. Our evaluations demonstrate that DECODE retrieves functionally similar drugs in zero-shot settings with over 20% relative improvement over chemical baselines in mechanism-of-action (MOA) prediction. Furthermore, the framework achieves a 6-fold increase in hit rates for novel anti-cancer agents during external validation. Availability and implementation: The codes and datasets of DECODE are available at https://github.com/lian-xiao/DECODE.

Modern neuroscience has accumulated extensive evidence on perception, memory, prediction, valuation, and consciousness, yet still lacks an explicit operational architecture capable of integrating these phenomena within a unified computational framework. Existing theories address specific aspects of neural function: predictive coding and active inference emphasize hierarchical inference and prediction error minimization; engram theories explain memory through distributed cell assemblies; neuromodulatory accounts focus on value-dependent regulation of plasticity and behaviour; and global workspace or large-scale network models investigate mechanisms underlying conscious access. Despite their explanatory power, these approaches remain only partially integrated at the architectural level. This work introduces DIME (Detect-Integrate-Mark-Execute), a neural architecture organizing perception, memory, valuation, and conscious access within a common operational cycle. The framework includes four interacting components: engrams, distributed recurrent neural structures supporting multiple activation trajectories; execution threads, spatiotemporal trajectories implementing neural processes; marker systems, neuromodulatory and limbic mechanisms regulating gain, plasticity, and trajectory selection; and hyperengrams, large-scale integrative states associated with operational conscious access. The framework is consistent with empirical evidence from hippocampal indexing, recurrent cortical processing, replay phenomena, large-scale network integration, and neuromodulatory regulation. Formulated at an abstract computational level, DIME may also inform artificial intelligence and robotics by providing an architectural template in which representation, valuation, and temporal sequencing emerge from a unified mechanism. An extended theoretical exposition is available in a companion monograph on Zenodo.

Intracranial language brain-computer interfaces (BCIs) are a promising route for restoring communication in people with severe motor and speech impairments, but clinical translation remains limited by fragmented evidence and unresolved design trade-offs across neuroscience, hardware, algorithm, evaluation, and clinical deployment. This review synthesizes progress in neural mechanisms of overt, mimed, and imagined speech; decision-oriented hardware comparisons of microelectrode array (MEA), electrocorticography (ECoG), and stereotactic electroencephalography (SEEG) recording modalities; experiment design for cross-subject and multilingual generalization; and neural decoding advances spanning sequence models, transformers, articulatory intermediates, and language-prior-assisted frameworks. We highlight persistent bottlenecks, including weak cross-subject transfer, long-term non-stationarity and recalibration burden, heterogeneous and non-comparable evaluation practices, limited naturalistic expressivity (especially for tonal/logosyllabic languages), and low signal-to-noise ratio (SNR) of neural activity in covert speech decoding. Our contributions are threefold: (1) an end-to-end, decision-oriented synthesis linking neural representations to recording choices, experimental design, decoding model architectures, and translational constraints; (2) a structured framework organized around five coupled design questions, together with a unified evaluation framework and a cross-language/cross-task benchmark template integrating objective, perceptual, expressive, conversational, and longitudinal metrics; and (3) user-centered translational guidance covering agency-preserving shared control, verifiable performance priorities, and scenario-specific minimum viable system (MVP) profiles for reliability-first home communication versus fidelity-first conversational speech restoration.

Motivation: Receptor-binding proteins (RBPs) initiate viral infection and determine host specificity, serving as key targets for phage engineering and therapy. However, the identification of RBPs is complicated by their extreme sequence divergence, which often renders traditional homology-based alignment methods ineffective. While machine learning offers a promising alternative, such approaches struggle with severe class imbalance and the difficulty of selecting informative negative samples from heterogeneous tail proteins. Existing methods often fail to balance learning from these ``hard negatives'' while maintaining generalization. Results: We present SeekRBP, a sequence--structure framework that models negative sampling as a sequential decision-making problem. By employing a multi-armed bandit strategy, SeekRBP dynamically prioritizes informative non-RBP sequences based on real-time training feedback, complemented by a multimodal fusion of protein language and structural embeddings. Benchmarking demonstrates that SeekRBP consistently outperforms static sampling strategies. Furthermore, a case study on Vibrio phages validates that SeekRBP effectively identifies RBPs to improve host prediction, highlighting its potential for large-scale annotation and synthetic biology applications.

From 1/f noise to neuronal avalanches, evidence of scaling in brain activity has been increasingly linked to tuning to or near criticality. The concept of scaling is intimately related to the renormalization group (RG), in essence providing coarse-grained, simplified descriptions that generalize to classes of diverse physical systems. Following the RG idea, scaling laws have been reported in populations of spiking neurons at microscopic scales. Whether similar scaling principles govern large-scale neural activity in the human brain and how they relate to underlying neural physiology remains unresolved. Here, we analyze large-scale electrophysiological recordings (MEG) of human resting-state brain activity and apply a RG-inspired coarse-graining approach to track collective neural dynamics across spatial scales. We find that multiple observables exhibit robust scale-invariant behavior under coarse-graining: activity variance and correlations grow according to power laws, covariance eigenspectra follow a characteristic scaling relation, and neuronal avalanche statistics remain invariant. Using an analytically tractable neural network model, we show that the observed scaling signatures arise when the system operates slightly below criticality, and that the scaling exponents depend on the excitation-inhibition balance. These findings demonstrate that RG-inspired scaling analysis can uncover signatures of critical dynamics in non-invasive human electrophysiology and suggest a principled route toward estimating excitation-inhibition balance from large-scale brain recordings.

Men who have sex with men (MSM) remain disproportionately affected by HIV, yet optimizing Pre-exposure Prophylaxis (PrEP) distribution remains a public health challenge. Current guidelines and most modelling studies do not incorporate sociodemographic or network-level factors that shape transmission. While network reconstruction from egocentric data has been studied, the relative importance of demographic mixing dimensions remains uncertain. Using data from 4,667 MSM participants, we show that uncertainty in network reconstruction from egocentric survey data - specifically whether assortativity by age or race is incorporated - affects simulated HIV prevalence under the same observed PrEP uptake. We simulate HIV transmission over 50 years across this structural space and evaluate whether empirically observed uptake reaches transmission-critical network positions. Network structure strongly influences outcomes: assortative by degree networks show 17% lower equilibrium prevalence due to hub isolation within communities. Targeted PrEP strategies based on degree or k-shell centrality achieved the highest prevalence reductions, particularly in assortative by age and race networks where hubs bridge demographic groups. PrEP uptake from data is suboptimal in assortative by age and race networks, underperforming compared with network-based strategies. Results demonstrate that uncertainty in network reconstruction affects intervention design and highlight the need for robust prevention strategies under structural ambiguity.

Background: The surge in single-cell omics data exposes limitations in traditional, manually defined analysis workflows. AI agents offer a paradigm shift, enabling adaptive planning, executable code generation, traceable decisions, and real-time knowledge fusion. However, the lack of a comprehensive benchmark critically hinders progress. Results: We introduce a novel benchmarking evaluation system to rigorously assess agent capabilities in single-cell omics analysis. This system comprises: a unified platform compatible with diverse agent frameworks and LLMs; multidimensional metrics assessing cognitive program synthesis, collaboration, execution efficiency, bioinformatics knowledge integration, and task completion quality; and 50 diverse real-world single-cell omics analysis tasks spanning multi-omics, species, and sequencing technologies. Our evaluation reveals that Grok3-beta achieves state-of-the-art performance among tested agent frameworks. Multi-agent frameworks significantly enhance collaboration and execution efficiency over single-agent approaches through specialized role division. Attribution analyses of agent capabilities identify that high-quality code generation is crucial for task success, and self-reflection has the most significant overall impact, followed by retrieval-augmented generation (RAG) and planning. Conclusions: This work highlights persistent challenges in code generation, long-context handling, and context-aware knowledge retrieval, providing a critical empirical foundation and best practices for developing robust AI agents in computational biology.

Traditional artificial neural networks consist of nodes with non-oscillatory dynamics. Biological neural networks, on the other hand, consist of oscillatory components embedded in an oscillatory environment. Motivated by this feature of biological neurons, we describe a neural network framework with explicit damped, oscillatory node dynamics. We express the oscillatory dynamics using two history dependent terms to connect these dynamics with standard recurrent neural network formulations, apply physical constraints from observed brain dynamics to choose the oscillatory frequencies, and stationary constraints to reduce the number of free parameters. We then optimize and illustrate network performance by classifying hand-written digits and simulated neuronal spike train activity and show that these oscillatory network elements support accurate classification with few trainable parameters. Choosing oscillator frequencies according to a proposed theory for brain rhythms improves classification accuracy compared to alternative frequency configurations and compared to standard recurrent neural network frameworks with comparable numbers of parameters. Compared to existing approaches, the proposed resonant recurrent network (RRN) utilizes oscillatory dynamics expressed as a straightforward extension of standard recurrent neural networks, produces interpretable features for classification, and performs well with few parameters when oscillator frequencies follow a configuration observed in vivo. We propose that RRNs may serve as efficient, biologically inspired building blocks to achieve complex goals in biological and artificial neural networks.

The cognitive state of mind concerning a range of choices to be made can effectively be modelled in terms of an element of a high-dimensional Hilbert space. The dynamics of the state of mind resulting form information acquisition is characterised by the von Neumann-Lüders projection postulate of quantum theory. This is shown to give rise to an uncertainty-minimising dynamical behaviour equivalent to the Bayesian updating, hence providing an alternative approach to characterising the dynamics of cognitive state that is consistent with the free energy principle in brain science. The quantum formalism however goes beyond the range of applicability of classical reasoning in explaining cognitive behaviours, thus opens up new and intriguing possibilities.

Turing instabilities of reaction-diffusion systems can only arise if the diffusivities of the chemical species are sufficiently different. This threshold is unphysical in most systems with $N=2$ diffusing species, forcing experimental realizations of the instability to rely on fluctuations or additional nondiffusing species. Here we ask whether this diffusive threshold lowers for $N>2$ to allow "true" Turing instabilities. Inspired by May's analysis of the stability of random ecological communities, we analyze the probability distribution of the diffusive threshold in reaction-diffusion systems defined by random matrices describing linearized dynamics near a homogeneous fixed point. In the numerically tractable cases $N\leqslant 6$, we find that the diffusive threshold becomes more likely to be smaller and physical as $N$ increases and that most of these many-species instabilities cannot be described by reduced models with fewer species.

Understanding cellular machinery requires atomic-scale reconstruction of large biomolecular assemblies. However, predicting the structures of these systems has been constrained by hardware memory requirements of models like AlphaFold 3, imposing a practical ceiling of a few thousand residues that can be processed on a single GPU. Here we present NVIDIA BioNeMo Fold-CP, a context parallelism framework that overcomes this barrier by distributing the inference and training pipelines of co-folding models across multiple GPUs. We use the Boltz models as open source reference architectures and implement custom multidimensional primitives that efficiently parallelize both the dense triangular updates and the irregular, data-dependent pattern of window-batched local attention. Our approach achieves efficient memory scaling; for an N-token input distributed across P GPUs, per-device memory scales as $O(N^2/P)$, enabling the structure prediction of assemblies exceeding 30,000 residues on 64 NVIDIA B300 GPUs. We demonstrate the scientific utility of this approach through successful developer use cases: Fold-CP enabled the scoring of over 90% of Comprehensive Resource of Mammalian protein complexes (CORUM) database, as well as folding of disease-relevant PI4KA lipid kinase complex bound to an intrinsically disordered region without cropping. By providing a scalable pathway for modeling massive systems with full global context, Fold-CP represents a significant step toward the realization of a virtual cell.

During the COVID-19 crisis, policymakers have implemented "social bubble" merging strategies, which allowed people from different households to meet and interact. Although these measures can mitigate the negative effects of extreme isolation, they also introduce additional contacts that may facilitate disease spread. As a result, several modeling studies have explored the epidemiological impact of different household-merging strategies, in which the selection of households to be merged is guided by specific demographic criteria, such as household size or the age composition of their members. Here we investigate an alternative pairing strategy in which households are merged according to the number of economically active (working) members. We develop a mathematical model of household networks using real demographic data from multiple regions around the world, and simulate a lockdown scenario in which only economically active individuals can leave their households, while the remaining non-working members stay indoors. By using numerical simulations and the generating function technique, we then estimate the epidemic risk for different household merging strategies. We found that merging strategies based on the number of working members can keep epidemic risk at similar levels as those based on household size. Moreover, the worker-based approach allows significantly more people to form larger social bubbles, exceeding 40\% of the population in some countries. We found that merging households with at most one worker provides the best balance between controlling epidemic risk and addressing people's need for social contact.

Autonomous LLM agents require structured long-term memory, yet current "append-and-evolve" systems like A-MEM face O(N^2) write-latency and excessive token costs. We introduce D-MEM (Dopamine-Gated Agentic Memory), a biologically inspired architecture that decouples short-term interaction from cognitive restructuring via a Fast/Slow routing system based on Reward Prediction Error (RPE). A lightweight Critic Router evaluates stimuli for Surprise and Utility. Routine, low-RPE inputs are bypassed or cached in an O(1) fast-access buffer. Conversely, high-RPE inputs, such as factual contradictions or preference shifts, trigger a "dopamine" signal, activating the O(N) memory evolution pipeline to reshape the agent's knowledge graph. To evaluate performance under realistic conditions, we introduce the LoCoMo-Noise benchmark, which injects controlled conversational noise into long-term sessions. Evaluations demonstrate that D-MEM reduces token consumption by over 80%, eliminates O(N^2) bottlenecks, and outperforms baselines in multi-hop reasoning and adversarial resilience. By selectively gating cognitive restructuring, D-MEM provides a scalable, cost-efficient foundation for lifelong agentic memory.

Carbapenem-resistant Pseudomonas aeruginosa (CRPA) and Acinetobacter baumannii (CRAB) represent a major clinical and epidemiological challenge and pose a growing threat to public health and the environment. Accordingly, CRPA and CRAB were investigated in hospital wastewater (HWW) collected during winter and summer 2024 from 64 healthcare facilities across all 16 Polish voivodeships. To our knowledge, this study constitutes the first nationwide, large-scale assessment in Poland of carbapenem resistance in these high-risk pathogens in hospital wastewater. The study aimed to determine the prevalence of carbapenem-resistant bacteria (CRB) in HWW discharged into the public sewer system and municipal wastewater treatment plants (WWTPs). In addition, associations between CRB prevalence, hospital geographic location, and sampling season were analyzed to identify spatial and temporal patterns of carbapenem resistance (CR). Carbapenem-resistant P. aeruginosa were predominant in all studied regions. Carbapenem-resistant A. baumannii were identified in a smaller percentage of samples and were characterized by greater genotypic diversity. The ERIC-PCR assay confirmed the presence of both closely related strains and unique genetic profiles, which suggests that CRB emissions into the environment have a complex character. The statistical analysis revealed significant relationships between CRB counts, the physicochemical parameters of HWW, and antibiotic concentrations in HWW samples. In addition, the tested samples harbored many antibiotic resistance genes (ARGs), which confirms that HWW is a significant reservoir of mobile genetic elements (MGEs) involved in the spread of antibiotic resistance. The results of the study indicate that HWW should be rigorously monitored and managed to minimize risks to public health and environment.

We present ScienceClaw + Infinite, a framework for autonomous scientific investigation in which independent agents conduct research without central coordination, and any contributor can deploy new agents into a shared ecosystem. The system is built around three components: an extensible registry of over 300 interoperable scientific skills, an artifact layer that preserves full computational lineage as a directed acyclic graph (DAG), and a structured platform for agent-based scientific discourse with provenance-aware governance. Agents select and chain tools based on their scientific profiles, produce immutable artifacts with typed metadata and parent lineage, and broadcast unsatisfied information needs to a shared global index. The ArtifactReactor enables plannerless coordination: peer agents discover and fulfill open needs through pressure-based scoring, while schema-overlap matching triggers multi-parent synthesis across independent analyses. An autonomous mutation layer actively prunes the expanding artifact DAG to resolve conflicting or redundant workflows, while persistent memory allows agents to continuously build upon complex epistemic states across multiple cycles. Infinite converts these outputs into auditable scientific records through structured posts, provenance views, and machine-readable discourse relations, with community feedback steering subsequent investigation cycles. Across four autonomous investigations, peptide design for the somatostatin receptor SSTR2, lightweight impact-resistant ceramic screening, cross-domain resonance bridging biology, materials, and music, and formal analogy construction between urban morphology and grain-boundary evolution, the framework demonstrates heterogeneous tool chaining, emergent convergence among independently operating agents, and traceable reasoning from raw computation to published finding.

Many disciplines need quantitative models that synthesize experimental data across multiple instances of the same general system. For example, neuroscientists must combine data from the brains of many individual animals to understand the species' brain in general. However, typical machine learning models treat one system instance at a time. Here we introduce a machine learning framework, deep probabilistic model synthesis (DPMS), that leverages system properties auxiliary to the model to combine data across system instances. DPMS specifically uses variational inference to learn a conditional prior distribution and instance-specific posterior distributions over model parameters that respectively tie together the system instances and capture their unique structure. DPMS can synthesize a wide variety of model classes, such as those for regression, classification, and dimensionality reduction, and we demonstrate its ability to improve upon single-instance models on synthetic data and whole-brain neural activity data from larval zebrafish.

Gene regulatory networks exhibit hierarchical organization across scales; capturing this structure mathematically requires a metric that distinguishes regulatory influence at each level. We show that the ultrametric of the $p$-adic integers $\mathbb{Z}_p$ -- whose self-similar nested-ball structure is a natural fractal encoding of multi-scale organization -- provides such a framework. Embedding the $N$-gene state space into $\mathbb{Z}_p$ and working over the complete, algebraically closed field $\mathbb{C}_p$, we prove the existence of rational functions that interpret the discrete dynamics and construct hierarchical approximations at each resolution level. These constructions yield a stability measure $μ$ -- aggregating how the dynamics contracts or expands across resolution levels -- and a ball-level classification of fixed points -- contracting, expanding, or isometric -- extending the attracting/repelling/indifferent trichotomy of non-Archimedean dynamics from points to balls. A key result is that $μ$ and the classification, although their definition and dynamical meaning require the analytical tools of $\mathbb{C}_p$, are fully determined by the discrete data. Minimizing $μ$ over all $N!$ gene orderings defines an optimal regulatory hierarchy; for the Arabidopsis thaliana floral development network ($N=13$, $p=2$), a $μ$-minimizing ordering places known master regulators -- UFO, EMF1, LFY, TFL1 -- in the leading positions and recovers the accepted developmental hierarchy without biological input beyond the transition map.

Vision foundation models trained with self-supervised objectives achieve strong performance across diverse tasks and exhibit emergent object segmentation properties. However, their alignment with human object perception remains poorly understood. Here, we introduce a behavioral benchmark in which participants make same/different object judgments for dot pairs on naturalistic scenes, scaling up a classical psychophysics paradigm to over 1000 trials. We test a diverse set of vision models using a simple readout from their representations to predict subjects' reaction times. We observe a steady improvement across model generations, with both architecture and training objective contributing to alignment, and transformer-based models trained with the DINO self-supervised objective showing the strongest performance. To investigate the source of this improvement, we propose a novel metric to quantify the object-centric component of representations by measuring patch similarity within and between objects. Across models, stronger object-centric structure predicts human segmentation behavior more accurately. We further show that matching the Gram matrix of supervised transformer models, capturing similarity structure across image patches, with that of a self-supervised model through distillation improves their alignment with human behavior, converging with the prior finding that Gram anchoring improves DINOv3's feature quality. Together, these results demonstrate that self-supervised vision models capture object structure in a behaviorally human-like manner, and that Gram matrix structure plays a role in driving perceptual alignment.

The blue shark (Prionace glauca) exhibits a striking dorsoventral color gradient, transitioning from vibrant blue dorsally to silver and white ventrally, a pattern widely interpreted as pelagic countershading. Despite its ecological significance, the physical basis of this coloration remains unresolved. Here we show that this color system does not arise from dermal chromatophores, as in most vertebrates, but from a previously unrecognised photonic architecture housed within the pulp cavity of individual dermal denticles that cover the skin. Optical imaging reveals discrete color domains within denticle crowns, while external denticle morphology remains similar across color zones. Using spectroscopy, micro-computed tomography, histology, and correlative electron microscopy, we demonstrate that color variation is organized across coupled micro- and nanoscale architectures. In blue denticles, iridophores and melanophores form a densely packed tessellated reflector-absorber system within an expanded crown-restricted pulp cavity. Transition-zone denticles exhibit partial cellular layering, whereas white denticles lack melanophores and contain only reflective cells. At the nanoscale, ordered purine-crystal stacks generate narrowband blue reflection, whereas disordered assemblies produce broadband white scattering. Together, these results reveal denticles as mechanically protected optical "pixels" whose hierarchical cellular and nanocrystal organization generates the shark's countershaded coloration.

Current biological AI models lack interpretability -- their internal representations do not correspond to biological relationships that researchers can examine. Understanding gene regulation requires models whose learned structure can be directly interrogated to generate experimentally testable hypotheses. CDT-II mirrors the central dogma in its architecture -- DNA self-attention, RNA self-attention, and cross-attention for transcriptional control -- requiring only genomic embeddings and raw per-cell expression. Applied to K562 CRISPRi data with five genes held out entirely, CDT-II predicts perturbation effects (per-gene mean r = 0.84), recovers the GFI1B regulatory network (6.6-fold enrichment, P = 3.5 x 10^{-17}), and shows that cross-attention focuses on ENCODE regulatory elements including CTCF sites (mean 7.67x across 28 targets, P < 0.001). Gradient-based attribution accurately predicts downstream consequences of perturbing therapeutic targets (mean r = 0.82). Applied to TFRC, the target of the anti-TfR1 antibody PPMX-T003, gradient analysis identifies genes involved in erythrocyte structure, iron-dependent DNA synthesis, and oxidative stress -- pathways that align with anemia and reticulocyte decrease reported in Phase 1 trials and ferroptosis demonstrated in preclinical studies, without any clinical data as input, establishing CDT-II as an AI microscope that reveals clinically relevant regulatory structure from perturbation experiments alone.

Coral colonies exhibit complex, self-similar branching architectures shaped by biochemical interactions and environmental constraints. To model their growth and calcification dynamics, we propose a novel p-adic reaction-diffusion framework defined over p-adic ultrametric spaces. The model incorporates biologically grounded reactions involving calcium and bicarbonate ions, whose interplay drives the precipitation of calcium carbonate (CaCO3). Nonlocal diffusion is governed by the Vladimirov operator over the p-adic integers, naturally capturing the hierarchical geometry of branching coral structures. Discretization over p-adic balls yields a high-dimensional nonlinear ODE system, which we solve numerically to examine how environmental and kinetic parameters, particularly CO2 concentration, influence morphogenetic outcomes. The resulting simulations reproduce structurally diverse and biologically plausible branching patterns. This approach bridges non-Archimedean analysis with morphogenesis modeling and provides a mathematically rigorous framework for investigating hierarchical structure formation in developmental biology.

Glioblastoma recurrence is largely driven by diffuse infiltration beyond radiologically visible tumor margins, yet standard radiotherapy, the mainstay of glioblastoma treatment, relies on uniform expansions that ignore patient-specific biological and anatomical factors. While computational models promise to map this invisible growth and guide personalized treatment planning, their clinical translation is hindered by the lack of standardized, large-scale benchmarking and reproducible validation workflows. To bridge this gap, we present PREDICT-GBM, a comprehensive open-source platform that integrates a curated, longitudinal, multi-center dataset of 243 patients with a standardized evaluation pipeline, and fuels model development and validation. We demonstrate PREDICT-GBM's potential by training and benchmarking a novel U-Net-based recurrence prediction model against state-of-the-art biophysical and data-driven methods. Our results show that both biophysical and deep-learning approaches significantly outperform standard-of-care protocols in predicting future recurrence sites while maintaining iso-volumetric treatment constraints. Notably, our U-Net model achieved a superior coverage of enhancing recurrence (79.37 +/- 2.08 %), markedly surpassing the standard-of-care (paired Wilcoxon signed-rank test, p = 0.0000057). Furthermore, the biophysical model GliODIL reached 78.91 +/- 2.08 % (p = 0.00045), validating the platform's ability to compare diverse modeling paradigms. By providing the first rigorous, reproducible ecosystem for model training and validation, PREDICT-GBM eliminates a major bottleneck for personalized, computationally guided radiotherapy. This work establishes a new standard for developing computationally guided, personalized radiotherapy, with the platform, models, and data openly available at github.com/BrainLesion/PredictGBM

Neurons communicate with downstream systems via sparse and incredibly brief electrical pulses, or spikes. Using these events, they control various targets such as neuromuscular units, neurosecretory systems, and other neurons in connected circuits. This gave rise to the idea of spiking neurons as controllers, in which spikes are the control signal. Using instantaneous events directly as the control inputs, also called `impulse control', is challenging as it does not scale well to larger networks and has low analytical tractability. Therefore, current spiking control usually relies on filtering the spike signal to approximate analog control. This ultimately means spiking neural networks (SNNs) have to output a continuous control signal, necessitating continuous energy input into downstream systems. Here, we circumvent the need for rate-based representations, providing a scalable method for task-specific spiking control with sparse neural activity. In doing so, we take inspiration from both optimal control and neuroscience theory, and define a spiking rule where spikes are only emitted if they bring a dynamical system closer to a target. From this principle, we derive the required connectivity for an SNN, and show that it can successfully control linear systems. We show that for physically constrained systems, predictive control is required, and the control signal ends up exploiting the passive dynamics of the downstream system to reach a target. Finally, we show that the control method scales to both high-dimensional networks and systems. Importantly, in all cases, we maintain a closed-form mathematical derivation of the network connectivity, the network dynamics and the control objective. This work advances the understanding of SNNs as biologically-inspired controllers, providing insight into how real neurons could exert control, and enabling applications in neuromorphic hardware design.

Previous studies have compared neural activities in the visual cortex to representations in deep neural networks trained on image classification. Interestingly, while some suggest that their representations are highly similar, others argued the opposite. Here, we propose a new approach to characterize the similarity of the decision strategies of two observers (models or brains) using decision variable correlation (DVC). DVC quantifies the image-by-image correlation between the decoded decisions based on the internal neural representations in a classification task. Thus, it can capture task-relevant information rather than general representational alignment. We evaluate DVC using monkey V4/IT recordings and network models trained on image classification tasks. We find that model-model similarity is comparable to monkey-monkey similarity, whereas model-monkey similarity is consistently lower. Strikingly, DVC decreases with increasing network performance on ImageNet-1k. Adversarial training does not improve model-monkey similarity in task-relevant dimensions assessed using DVC, although it markedly increases the model-model similarity. Similarly, pre-training on larger datasets does not improve model-monkey similarity. These results suggest a divergence between the task-relevant representations in monkey V4/IT and those learned by models trained on image classification tasks.