Navigation up a sensory gradient is one of the simplest behaviours, and the simplest strategy is run and tumble. But some organisms use other strategies, such as reversing direction or turning by some angle. Here we ask what drives the choice of strategy, which we frame as maximising up-gradient speed using a given amount of sensory information per unit time. We find that, without directional information on which way to turn, behavioural strategies which make sudden turns perform better than gradual steering. We see various transitions where a different strategy becomes optimal, such as a switch from reversing direction to fully re-orienting tumbles as more information becomes available. And, among more complex re-orientation strategies, we show that discrete turn angles are best, and see transitions in how many such angles the optimal strategy employs.

Simulation code for conventional supercomputers serves as a reference for neuromorphic computing systems. The present bottleneck of distributed large-scale spiking neuronal network simulations is the communication between compute nodes. Communication speed seems limited by the interconnect between the nodes and the software library orchestrating the data transfer. Profiling reveals, however, that the variability of the time required by the compute nodes between communication calls is large. The bottleneck is in fact the waiting time for the slowest node. A statistical model explains total simulation time on the basis of the distribution of computation times between communication calls. A fundamental cure is to avoid communication calls because this requires fewer synchronizations and reduces the variability of computation times across compute nodes. The organization of the mammalian brain into areas lends itself to such an optimization strategy. Connections between neurons within an area have short delays, but the delays of the long-range connections across areas are an order of magnitude longer. This suggests a structure-aware mapping of areas to compute nodes allowing for a partition into more frequent communication between nodes simulating a particular area and less frequent global communication. We demonstrate a substantial performance gain on a real-world example. This work proposes a local-global hybrid communication architecture for large-scale neuronal network simulations as a first step in mapping the structure of the brain to the structure of a supercomputer. It challenges the long-standing belief that the bottleneck of simulation is synchronization inherent in the collective calls of standard communication libraries. We provide guidelines for the energy efficient simulation of neuronal networks on conventional computing systems and raise the bar for neuromorphic systems.

Generalizable protein function prediction is increasingly constrained by the growing mismatch between exponentially expanding sequences of environmental proteins and the comparatively slow accumulation of experimentally verified functional data. Active learning offers a promising path forward for accelerating biological function prediction, by selecting the most informative proteins to experimentally annotate for data-efficient training, yet its potential remains largely unexplored. We introduce HATTER (Human-in-the-loop Adaptive Toolkit for Transferable Enzyme Representations), a modular framework that integrates multiple active learning strategies with human-in-the-loop experimental annotation to efficiently fine tune function prediction models. We compare active learning training to standard supervised training for biological enzyme function prediction, demonstrating that active learning achieves performance comparable to standard training across diverse protein sequence evaluation datasets while requiring fewer model updates, processing less data, and substantially reducing computational cost. Interestingly, point-based uncertainty sampling methods like entropy or margin sampling perform as well or better than more complex acquisition functions such as bayesian sampling or BALD, highlighting the relative importance of sequence diversity in training datasets and model architecture design. These results demonstrate that human-in-the-loop active learning can efficiently accelerate enzyme discovery, providing a flexible platform for adaptive, scalable, and expert-guided protein function prediction.

Dale's Principle has historically guided neuroscience research as a valuable rule of thumb, namely that all synapses on each neuron release the same set of neurotransmitters. Most existing Spiking Neuron Network models share this dichotomous assumption that neurons are either excitatory or inhibitory; however, recent experimental evidence points towards co-release mechanisms that violate this assumption. Here, we introduce a minimal model of "Bilingual" neurons violating Dale's principle that can exert both excitatory and inhibitory effects. We identify parameter regimes in which this architecture exhibits transitions between synchronous and asynchronous dynamics that differ quantitatively from those observed in a matched monolingual control architecture. We report distinct information-processing signatures both at the level of neurons and higher-order interactions between them near the phase transitions. These results suggest that the population of neurons violating Dales principle may provide an alternative mechanism for regulating large-scale oscillatory activity in neural circuits.

We propose a minimal mathematical framework to describe the go-or-grow dynamics of tumor cells comprising two phenotypically distinct populations. One population is migratory and undergoes linear diffusion, while the other proliferates in an oxygen-dependent manner. The local oxygen concentration governs transitions between these phenotypes. We then ask whether these two coupled phenotype-specific equations can be reduced to a single mixed-phenotype equation under cyclic hypoxia. We establish a connection between the minimal go-or-grow model with distinct phenotypic populations and a reduced model describing a single-cell population with oxygen-dependent diffusion and proliferation in the fast-phenotypic-switching regime. This theoretical reduction is validated through numerical simulations.

Implementations of symmetric positive definite (SPD) matrix-based neural networks for neural decoding remain fragmented across research codebases and Python packages. Existing implementations often employ ad hoc handling of manifold constraints and non-unified training setups, which hinders reproducibility and integration into modern deep-learning workflows. To address this gap, we introduce SPD Learn, a unified and modular Python package for geometric deep learning with SPD matrices. SPD Learn provides core SPD operators and neural-network layers, including numerically stable spectral operators, and enforces Stiefel/SPD constraints via trivialization-based parameterizations. This design enables standard backpropagation and optimization in unconstrained Euclidean spaces while producing manifold-constrained parameters by construction. The package also offers reference implementations of representative SPDNet-based models and interfaces with widely used brain computer interface/neuroimaging toolkits and modern machine-learning libraries (e.g., MOABB, Braindecode, Nilearn, and SKADA), facilitating reproducible benchmarking and practical deployment.

Morphogenesis involves complex shape changes of biological tissues. Yet, tissue shape changes depend on tissue rheology, which in turn arises from the interplay of large numbers of cells. Here, we link cell- and tissue-scale mechanics by constructing mean-field rheological relations for the vertex model. In contrast to past work in the field, we study a vertex model with an explicit viscous friction. We also include two different cellular mechanisms creating active, anisotropic stresses. Our mean-field model accounts for cell shape and the non-linear elastic and visco-plastic regimes. We validate our results by predicting the response to large-amplitude oscillatory shear. There are several vertex model variants, and comparing to results from the literature, we show that their rheology depends on a number of model details. Our approach should be sufficiently general to construct non-linear mean-field constitutive relations for any cell-based tissue model.

Branching processes are models used to describe populations that reproduce and die over time. In the classical setting, an individual's reproductive capacity remains constant throughout its lifetime. However, in real-world situations, reproductive capacity typically undergoes ageing - that is, after reaching a peak, it decreases over time. In this work, we study the influence of ageing on the behaviour of the process and how modifying its parameters, along with reproduction rates, affects the destiny of the process.

Herbivorous wild species constantly strive to optimize the trade-off between energy and nutrient intake and predation risk during foraging. This has led to the selection of several evolutionary traits -- such as diet, habitat selection, and behavior -- which are simultaneously shaped by the spatio-temporal variability of the habitat. Among camelid species, polygyny is a prevalent behavioral strategy that encompasses both mating and foraging activities. This group-level behavior has multiple interacting dimensions, contributing to an interesting ecological and evolutionary complexity. We developed an individual-based stochastic model in which camelid females transition between different familial groups in response to their environmental conditions, aiming to maximize individual fitness. Our results indicate that the behavioral strategy of individual females can shape, by itself, emergent population-level properties, including group size and fitness distribution. Furthermore, these properties are modulated, in a non-additive manner, by other factors such as population density, sex ratio and system heterogeneity.

Generative artificial intelligence models learn probability distributions from data and produce novel samples that capture the salient properties of their training sets. Proteins are particularly attractive for such approaches given their abundant data and the versatility of their representations, ranging from sequences to structures and functions. This versatility has motivated the rapid development of generative models for protein design, enabling the generation of functional proteins and enzymes with unprecedented success. However, because these models mirror their training distribution, they tend to sample from its most probable modes, while low-probability regions, often encoding valuable properties, remain underexplored. To address this challenge, recent work has proposed strategies for steering generative models toward user-specified properties. In this review, we survey and categorize these strategies, distinguishing approaches that modify model parameters, such as reinforcement learning or supervised fine-tuning, from those that keep the model's parameters fixed, including conditional generation, retrieval-augmented strategies, Bayesian guidance, and tailored sampling methods. Together, these developments are beginning to enable the steering of generative models toward proteins with desired properties.

Single-cell RNA-seq provides detailed molecular snapshots of individual cells but is notoriously noisy. Variability stems from biological differences and technical factors, such as amplification bias and limited RNA capture efficiency, making it challenging to adapt computational pipelines to heterogeneous datasets or evolving technologies. As a result, most studies still rely on principal component analysis (PCA) for dimensionality reduction, valued for its interpretability and robustness, in spite of its known bias in high dimensions. Here, we improve upon PCA with a Random Matrix Theory (RMT)-based approach that guides the inference of sparse principal components using existing sparse PCA algorithms. We first introduce a novel biwhitening algorithm which self-consistently estimates the magnitude of transcriptomic noise affecting each gene in individual cells, without assuming a specific noise distribution. This enables the use of an RMT-based criterion to automatically select the sparsity level, rendering sparse PCA nearly parameter-free. Our mathematically grounded approach retains the interpretability of PCA while enabling robust, hands-off inference of sparse principal components. Across seven single-cell RNA-seq technologies and four sparse PCA algorithms, we show that this method systematically improves the reconstruction of the principal subspace and consistently outperforms PCA-, autoencoder-, and diffusion-based methods in cell-type classification tasks.

Protein language models (PLMs) learn probability distributions over natural protein sequences. By learning from hundreds of millions of natural protein sequences, protein understanding and design capabilities emerge. Recent works have shown that scaling these models improves structure prediction, but does not seem to improve mutation understanding and representation quality for protein function prediction. We introduce PoET-2, a multimodal, retrieval-augmented protein foundation model that incorporates in-context learning of family-specific evolutionary constraints with optional structure conditioning to learn generative distributions over protein sequences. PoET-2 uses a hierarchical transformer encoder that is equivariant to sequence context ordering and a dual decoder architecture with both causal and masked language modeling objectives, allowing PoET-2 to operate in both fully generative and bidirectional representation learning modes. PoET-2 achieves state-of-the-art performance on zero-shot variant effect prediction, excelling at scoring variants with multiple mutations and challenging indel mutations. In supervised settings, PoET-2 embeddings outperform previous methods for learning sequence-function relationships, especially with small datasets. This work highlights the benefits of combining retrieval augmentation with multimodal, family-centric modeling for advancing protein foundation models.

Understanding how 5' untranslated regions (5'UTRs) regulate mRNA translation is critical for controlling protein expression and designing effective therapeutic mRNAs. While recent deep learning models have shown promise in predicting translational efficiency from 5'UTR sequences, most are constrained by fixed input lengths and limited interpretability. We introduce UTR-STCNet, a Transformer-based architecture for flexible and biologically grounded modeling of variable-length 5'UTRs. UTR-STCNet integrates a Saliency-Aware Token Clustering (SATC) module that iteratively aggregates nucleotide tokens into multi-scale, semantically meaningful units based on saliency scores. A Saliency-Guided Transformer (SGT) block then captures both local and distal regulatory dependencies using a lightweight attention mechanism. This combined architecture achieves efficient and interpretable modeling without input truncation or increased computational cost. Evaluated across three benchmark datasets, UTR-STCNet consistently outperforms state-of-the-art baselines in predicting mean ribosome load (MRL), a key proxy for translational efficiency. Moreover, the model recovers known functional elements such as upstream AUGs and Kozak motifs, highlighting its potential for mechanistic insight into translation regulation.

Animals flexibly recombine a finite set of core motor motifs to meet diverse task demands, but existing behavior segmentation methods oversimplify this process by imposing discrete syllables under restrictive generative assumptions. To better capture the continuous structure of behavior generation, we introduce motif-based continuous dynamics (MCD) discovery, a framework that (1) uncovers interpretable motif sets as latent basis functions of behavior by leveraging representations of behavioral transition structure, and (2) models behavioral dynamics as continuously evolving mixtures of these motifs. We validate MCD on a multi-task gridworld, a labyrinth navigation task, and freely moving animal behavior. Across settings, it identifies reusable motif components, captures continuous compositional dynamics, and generates realistic trajectories beyond the capabilities of traditional discrete segmentation models. By providing a generative account of how complex animal behaviors emerge from dynamic combinations of fundamental motor motifs, our approach advances the quantitative study of natural behavior.

The spreading dynamics of infectious diseases is influenced by individual behaviours, which are in turn affected by the level of awareness about the epidemic. Modelling the co-evolution of disease transmission and behavioural changes within a population enables better understanding, prediction and control of epidemics. Here, our primary goal is to provide an overview of the most popular modelling approaches, ranging from compartmental mean-field to agent-based models, with a particular focus on how behavioural factors are incorporated into epidemic dynamics. We classify modelling approaches based on the fundamental conceptual distinction between models of behaviours and models of behavioural determinants (such as awareness, beliefs, opinions, or trust); in particular, we observe that most studies model and interpret the variables related to individual responses either as behaviours or as determinants, with the implicit assumption that they correlate linearly. Based on preliminary empirical observations, we then challenge this assumption by analysing a recent dataset about time series of social indicators, collected during the COVID-19 pandemic. We examine the case study of Italian regions and we discover that behavioural responses are poorly explained by awareness, beliefs or trust, thereby calling for a careful interpretation of the modelling assumptions and for the development of further models, which fully account for the inherent complexity of individual responses and human behaviours.

There is a morphodynamic component to synaptic learning by which changes in dendritic (postsynaptic) spine head size are associated with the strengthening or weakening of the synaptic connection between two neurons. The membrane shape and size dynamics is sculpted by the growth dynamics of the enclosed actin cytoskeleton. We use Dynamical Graph Grammars (DGGs) governing dynamic labelled graphs embedded in two dimensions to model networks of actin filaments and the enclosing membrane in spine head morphology. We demonstrate the flexibility and extensibility of the framework by encoding detailed biophysical as well as biochemical models, obeying constraints of invariance and conservation, in DGG rule sets. From graph-local energy functions for cytoskeleton actin interacting and membrane, we specialize dissipative stochastic dynamics to an exhaustive collection of graph-local neighborhood types for the rule left hand sides. Extensively simulating the resulting model delineates effects of four actin-binding proteins, and their epistatic relationships, on morphology.

Graphs provide a powerful framework for modeling complex systems, but their structural variability poses significant challenges for analysis and classification. To address these challenges, we introduce GAUDI (Graph Autoencoder Uncovering Descriptive Information), a novel unsupervised geometric deep learning framework designed to capture both local details and global structure. GAUDI employs an innovative hourglass architecture with hierarchical pooling and upsampling layers linked through skip connections, which preserve essential connectivity information throughout the encoding-decoding process. Even though identical or highly similar underlying parameters describing a system's state can lead to significant variability in graph realizations, GAUDI consistently maps them into nearby regions of a structured and continuous latent space, effectively disentangling invariant process-level features from stochastic noise. We demonstrate GAUDI's versatility across multiple applications, including small-world networks modeling, characterization of protein assemblies from super-resolution microscopy, analysis of collective motion in the Vicsek model, and identification of age-related changes in brain connectivity. Comparison with related approaches highlights GAUDI's superior performance in analyzing complex graphs, providing new insights into emergent phenomena across diverse scientific domains.

Quality molecular representations are key to foundation model development in bio-medical research. Previous efforts have typically focused on a single representation or molecular view, which may have strengths or weaknesses on a given task. We develop Multi-view Molecular Embedding with Late Fusion (MMELON), an approach that integrates graph, image and text views in a foundation model setting and may be readily extended to additional representations. Single-view foundation models are each pre-trained on a dataset of up to 200M molecules. The multi-view model performs robustly, matching the performance of the highest-ranked single-view. It is validated on over 120 tasks, including molecular solubility, ADME properties, and activity against G Protein-Coupled receptors (GPCRs). We identify 33 GPCRs that are related to Alzheimer's disease and employ the multi-view model to select strong binders from a compound screen. Predictions are validated through structure-based modeling and identification of key binding motifs.

We propose a novel, tractable latent state inference scheme for Markov jump processes, for which exact inference is often intractable. Our approach is based on an entropic matching framework that can be embedded into the well-known expectation propagation algorithm. We demonstrate the effectiveness of our method by providing closed-form results for a simple family of approximate distributions and apply it to the general class of chemical reaction networks, which are a crucial tool for modeling in systems biology. Moreover, we derive closed-form expressions for point estimation of the underlying parameters using an approximate expectation maximization procedure. We evaluate our method across various chemical reaction networks and compare it to multiple baseline approaches, demonstrating superior performance in approximating the mean of the posterior process. Finally, we discuss the limitations of our method and potential avenues for future improvement, highlighting its promising direction for addressing complex continuous-time Bayesian inference problems.

We have developed URDME, a general software for simulation of stochastic reaction-diffusion processes on unstructured meshes. This allows for a more flexible handling of complicated geometries and curved boundaries compared to simulations on structured, cartesian meshes. The underlying algorithm is the next subvolume method, extended to unstructured meshes by obtaining jump coefficients from a finite element formulation of the corresponding macroscopic equation. This manual describes version 1.5 of the software. Refer to http://www.urdme.org for the latest updates.

While contemporary large language models (LLMs) are increasingly capable in isolation, there are still many difficult problems that lie beyond the abilities of a single LLM. For such tasks, there is still uncertainty about how best to take many LLMs as parts and combine them into a greater whole. This position paper argues that potential blueprints for designing such modular language agents can be found in the existing literature on cognitive models and artificial intelligence (AI) algorithms. To make this point clear, we formalize the idea of an agent template that specifies roles for individual LLMs and how their functionalities should be composed. We then survey a variety of existing language agents in the literature and highlight their underlying templates derived directly from cognitive models or AI algorithms. By highlighting these designs, we aim to call attention to agent templates inspired by cognitive science and AI as a powerful tool for developing effective, interpretable language agents.

The introduction of stochasticity into continuous ecological models frequently relies on phenomenological, diagonal diffusion terms that lack a rigorous microscopic basis. We demonstrate that this standard practice fundamentally misrepresents the geometry of demographic fluctuations. By deriving a stochastic Rosenzweig--MacArthur model directly from an integer-valued, Bernoulli-coupled continuous-time Markov chain, we isolate the exact diffusion covariance structure dictated by event stoichiometry. We mathematically prove that coupled predation--conversion events inherently generate a structurally negative predator--prey cross-covariance, exposing the severe mathematical and biological limitations of standard diagonal-noise approximations. Furthermore, we resolve a persistent ambiguity in stochastic population modeling by explicitly formalizing the bifurcation between open-domain formulations (for survival-conditioned interior dynamics) and absorbed formulations (for extinction-permitting dynamics). To rigorously support this distinction, we develop a tailored two-stage Lyapunov well-posedness architecture that separates non-explosion criteria from boundary-barrier positivity invariance. By bridging microscopic event stoichiometry with macroscopic boundary-degenerate diffusions, this work replaces ad hoc noise constructs with a definitive, mathematically exact template for covariance-consistent and boundary-aware ecological modeling.

Humans exhibit remarkable flexibility in abstract reasoning, and can rapidly learn and apply rules from sparse examples. To investigate the cognitive strategies underlying this ability, we introduce the Cognitive Abstraction and Reasoning Corpus (CogARC), a diverse human-adapted subset of the Abstraction and Reasoning Corpus (ARC) which was originally developed to benchmark abstract reasoning in artificial intelligence. Across two experiments, CogARC was administered to a total of 260 human participants who freely generated solutions to 75 abstract visual reasoning problems. Success required inferring input-output rules from a small number of examples to transform the test input into one correct test output. Participants' behavior was recorded at high temporal resolution, including example viewing, edit sequences, and multi-attempt submissions. Participants were generally successful (mean accuracy ~90% for experiment 1 (n=40), ~80% for experiment 2 (n=220) across problems), but performance varied widely across problems and participants. Harder problems elicited longer deliberation times and greater divergence in solution strategies. Over the course of the task, participants initiated responses more quickly but showed a slight decline in accuracy, suggesting increased familiarity with the task structure rather than improved rule-learning ability. Importantly, even incorrect solutions were often highly convergent, even when the problem-solving trajectories differed in length and smoothness. Some trajectories progressed directly and efficiently toward a stable outcome, whereas others involved extended exploration or partial restarts before converging. Together, these findings highlight CogARC as a rich behavioral environment for studying human abstract reasoning, providing insight into how people generalize, misgeneralize, and adapt their strategies under uncertainty.

Modern forward electrocardiogram (ECG) computational models rely on an accurate representation of the torso domain. The lead-field method enables fast ECG simulations while preserving full geometric fidelity. Achieving high anatomical accuracy in torso representation is, however, challenging in clinical practice, as imaging protocols are typically focused on the heart and often do not include the entire torso. In addition, the computational cost of the lead-field method scales linearly with the number of electrodes, limiting its applicability in high-density recording settings. To date, no existing approach simultaneously achieves high anatomical fidelity, low data requirements and computational efficiency. In this work, we propose a shape-informed surrogate model of the lead-field operator that serves as a drop-in replacement for the full-order model in forward ECG simulations. The proposed framework consists of two components: a geometry-encoding module that maps anatomical shapes into a low-dimensional latent space, and a geometry-conditioned neural surrogate that predicts lead-field gradients from spatial coordinates, electrode positions and latent codes. The proposed method achieves high accuracy in approximating lead fields both within the torso (mean angular error 5°) and inside the heart, resulting in highly accurate ECG simulations (relative mean squared error <2.5%. The surrogate consistently outperforms the widely used pseudo lead-field approximation while preserving negligible inference cost. Owing to its compact latent representation, the method does not require a fully detailed torso segmentation and can therefore be deployed in data-limited settings while preserving high-fidelity ECG simulations.

Cultures of neurons grown on multi-electrode arrays have become a common experimental preparation for investigating developing neural networks. Experiment and simulation have shown that these developing networks eventually exhibit bursting behavior in which the entire culture participates for short periods of time, with inter-burst intervals in which the network is comparatively quiescent. This paper extends previous simulation results by examining the spatiotemporal patterns of such bursting. We show that these bursts originate at a small number of network locations and propagate as waves of activity. We demonstrate that this type of activity does not require fine tuning of neuron or network parameters. We also examine how this activity changes during development and the dependence of such activity and its triggering on both local and global network properties.

Single-cell foundation models such as scGPT learn high-dimensional gene representations, but what biological knowledge these representations encode remains unclear. We systematically decode the geometric structure of scGPT internal representations through 63 iterations of automated hypothesis screening (183 hypotheses tested), revealing that the model organizes genes into a structured biological coordinate system rather than an opaque feature space. The dominant spectral axis separates genes by subcellular localization, with secreted proteins at one pole and cytosolic proteins at the other. Intermediate transformer layers transiently encode mitochondrial and ER compartments in a sequence that mirrors the cellular secretory pathway. Orthogonal axes encode protein-protein interaction networks with graded fidelity to experimentally measured interaction strength (Spearman rho = 1.000 across n = 5 STRING confidence quintiles, p = 0.017). In a compact six-dimensional spectral subspace, the model distinguishes transcription factors from their target genes (AUROC = 0.744, all 12 layers significant). Early layers preserve which specific genes regulate which targets, while deeper layers compress this into a coarser regulator versus regulated distinction. Repression edges are geometrically more prominent than activation edges, and B-cell master regulators BATF and BACH2 show convergence toward the B-cell identity anchor PAX5 across transformer depth. Cell-type marker genes cluster with high fidelity (AUROC = 0.851). Residual-stream geometry encodes biological structure complementary to attention patterns. These results indicate that biological transformers learn an interpretable internal model of cellular organization, with implications for regulatory network inference, drug target prioritization, and model auditing.

Mutational signature analysis has emerged as a powerful method for uncovering the underlying biological processes driving cancer development. However, the signature extraction process, typically performed using non-negative matrix factorization (NMF), often lacks reliability and clinical applicability. To address these limitations, several solutions have been introduced, including the use of neural networks to achieve more accurate estimates and probabilistic methods to better capture natural variation in the data. In this work, we introduce a Variational Autoencoder for Mutational Signatures (VAE-MS), a novel model that leverages both an asymmetric architecture and probabilistic methods for the extraction of mutational signatures. VAE-MS is compared to with three state-of-the-art models for mutational signature extraction: SigProfilerExtractor, the NMF-based gold standard; MUSE-XAE, an autoencoder that employs an asymmetric design without probabilistic components; and SigneR, a Bayesian NMF model, to illustrate the strength in combining a nonlinear extraction with a probabilistic model. In the ability to reconstruct input data and generalize to unseen data, models with probabilistic components (VAE-MS, SigneR) dramatically outperformed models without (SigProfilerExtractor, MUSE-XAE). The NMF-baed models (SigneR, SigProfilerExtractor) had the most accurate reconstructions in simulated data, while VAE-MS reconstructed more accurately on real cancer data. Upon evaluating the ability to extract signatures consistently, no model exhibited a clear advantage over the others. Software for VAE-MS is available at https://github.com/CLINDA-AAU/VAE-MS.

Accurate prediction of RNA-associated interactions is essential for understanding cellular regulation and advancing drug discovery. While Biological Large Language Models (BioLLMs) such as ESM-2 and RiNALMo provide powerful sequence representations, existing methods rely on static fusion strategies that fail to capture the dynamic, context-dependent nature of molecular binding. We introduce CrossLLM-Mamba, a novel framework that reformulates interaction prediction as a state-space alignment problem. By leveraging bidirectional Mamba encoders, our approach enables deep ``crosstalk'' between modality-specific embeddings through hidden state propagation, modeling interactions as dynamic sequence transitions rather than static feature overlaps. The framework maintains linear computational complexity, making it scalable to high-dimensional BioLLM embeddings. We further incorporate Gaussian noise injection and Focal Loss to enhance robustness against hard-negative samples. Comprehensive experiments across three interaction categories, RNA-protein, RNA-small molecule, and RNA-RNA demonstrate that CrossLLM-Mamba achieves state-of-the-art performance. On the RPI1460 benchmark, our model attains an MCC of 0.892, surpassing the previous best by 5.2\%. For binding affinity prediction, we achieve Pearson correlations exceeding 0.95 on riboswitch and repeat RNA subtypes. These results establish state-space modeling as a powerful paradigm for multi-modal biological interaction prediction.

Diffusion magnetic resonance imaging (dMRI) plays a vital role in both clinical diagnostics and neuroscience research. However, its inherently low signal-to-noise ratio (SNR), especially under high diffusion weighting, significantly degrades image quality and impairs downstream analysis. Recent self-supervised and unsupervised denoising methods offer a practical solution by enhancing image quality without requiring clean references. However, most of these methods do not explicitly account for the non-Gaussian noise characteristics commonly present in dMRI magnitude data during the supervised learning process, potentially leading to systematic bias and heteroscedastic variance, particularly under low-SNR conditions. To overcome this limitation, we introduce noise-corrected training objectives that explicitly model Rician statistics. Specifically, we propose two alternative loss functions: one derived from the first-order moment to remove mean bias, and another from the second-order moment to correct squared-signal bias. Both losses include adaptive weighting to account for variance heterogeneity and can be used without changing the network architecture. These objectives are instantiated in an image-specific, unsupervised Deep Image Prior (DIP) framework. Comprehensive experiments on simulated and in-vivo dMRI show that the proposed losses effectively reduce Rician bias and suppress noise fluctuations, yielding higher image quality and more reliable diffusion metrics than state-of-the-art denoising baselines. These results underscore the importance of bias- and variance-aware noise modeling for robust dMRI analysis under low-SNR conditions.

The brain anticipates future events using internal models that specify not only what will occur, but also when it will occur and with what probability. We refer to this joint specification of identity, timing, and likelihood as a complete prediction object. Existing computational models typically capture identity and timing separately, omit probability as an explicit representational dimension, or rely on biologically implausible global learning rules. Here we show that a single population of spiking neurons can acquire and flexibly maintain a complete prediction object through biologically grounded learning. We implemented a heterogeneous Izhikevich spiking reservoir with multiplexed readouts trained by an error-modulated, attention-gated three-factor Hebbian rule, and tested it on a task that independently manipulates event identity, latency, and probability. The network develops time-locked anticipatory activity whose amplitude scales with outcome probability and rapidly adapts when timing or probability statistics change. Identity and timing components self-organize into near-orthogonal readout subspaces within a shared neural population, demonstrating that multidimensional predictive structure can emerge without anatomical modularization or global error broadcast. Compared with least-squares-based approaches, local gated plasticity enables stable recalibration under nonstationary conditions. These results suggest that cortical mixed-selective populations, coupled with neuromodulator-gated synaptic plasticity, may be sufficient to jointly encode and update identity, timing, and probability within a single recurrent circuit. Flexible predictive cognition may therefore arise from generic population dynamics shaped by local learning rules rather than from specialized predictive modules.