We introduce Hyper Input Convex Neural Networks (HyCNNs), a novel neural network architecture designed for learning convex functions. HyCNNs combine the principles of Maxout networks with input convex neural networks (ICNNs) to create a neural network that is always convex in the input, theoretically capable of leveraging depth, and performs reliable when trained at scale compared to ICNNs. Concretely, we prove that HyCNNs require exponentially fewer parameters than ICNNs to approximate quadratic functions up to a given precision. Throughout a series of synthetic experiments, we demonstrate that HyCNNs outperform existing ICNNs and MLPs in terms of predictive performance for convex regression and interpolation tasks. We further apply HyCNNs to learn high-dimensional optimal transport maps for synthetic examples and for single-cell RNA sequencing data, where they oftentimes outperform ICNN-based neural optimal transport methods and other baselines across a wide range of settings.

It is a fundamental question in epidemiology to estimate, model and predict the growth rate of a pandemic. Analogously, analysing the diffusion of innovation, (fake) news, memes, and rumours is of key importance in the social sciences. The resulting epidemic growth curves can be classified according to their growth rates. These have been found to range from exponential to both faster super-exponential curves and slower subexponential or polynomial curves. Previous research has lacked a unified explanatory framework capable of accommodating super-exponential, (stretched) exponential, and polynomial growth patterns within the same contact network. In this paper we propose a simple agent-based network model that can capture all these phases. We provide such a framework by modelling how transmission rates depend on spatial distance and on individuals' numbers of contacts. By comparing the growth rate of spreading processes with or without degree-dependent and/or distance-dependent contact rates through data-driven and synthetic simulations on real and modelled networks with underlying geometry, we find evidence that even a 'sublinear presence' of these causes may cause a significant slow down of the growth rate on the same underlying network. We find that the growth rate is governed by a combination of three factors: geometry, the prevalence of weak ties, and superspreaders. We confirm our results with rigorous proofs in a theoretical model, using a spatial multiscale-argument in long-range heterogeneous first passage percolation. Our results give a plausible explanation of why the consecutive waves of a single pandemic can differ in their growth even if their spreading mechanisms are similar.

Cell invasion and spatial pattern formation are two distinct manifestations of cellular self-organisation in development, regeneration, and disease. Here, we develop and analyse a unified theoretical framework that links these two seemingly different behaviours within a single mechanistic model for adhesion-mediated self-organisation in growing cell populations. Using a multiscale analysis, we show that the balance between cell-cell adhesion, self-diffusion, and proliferation controls the emergence of distinct collective dynamics. We find that for weak adhesion, tissues invade through stable monotone fronts. As adhesion increases, invasion slows, fronts become unstable, leading to aggregates and spatial patterns emerging behind the advancing edge. In two spatial dimensions, these instabilities generate fingering morphologies reminiscent of dysregulated invasion in cancer. Crucially, we show that density-dependent regulation of adhesion suppresses these instabilities and restores cohesive tissue expansion. Together, our results identify adhesion strength and its regulation as key determinants of whether tissues invade cohesively or fragment into patterns, and provide a unified framework for understanding collective migration, morphogenesis, and dysregulated growth.

Boolean networks are powerful mathematical tools for modeling the qualitative dynamics of genetic regulation. Yet inferred models often generate spurious attractors that lack biological viability. In this paper, we propose a parsimonious computational framework to systematically refine Boolean network models by eliminating these non-biological asymptotic behaviors while strictly preserving known, biologically relevant attractors. Through an exhaustive exploration of local function substitutions, we generate a comprehensive set of candidate models. To identify the most biologically consistent networks, we implement an incremental pruning protocol that filters candidates based on structural interaction digraph similarity, attraction basin topological organization, trajectorial isomorphism, and the minimization of dynamical instability and frustration. We apply this methodology to a 9-node genetic control model of the osteogenesis regulation network. Our protocol effectively evaluates a syntactic search space of 51,138 potential networks, ultimately narrowing them down to a robust family of 6 parsimonious models that are fully compatible with current biological knowledge.

Epidemic forecasting has become an integral part of real-time infectious disease outbreak response. While collaborative ensembles composed of statistical and machine learning models have become the norm for real-time forecasting, standardized benchmark datasets for evaluating such methods are lacking. Further, there is limited understanding on performance of these methods for novel outbreaks with limited historical data. In this paper, we propose IDOBE, a curated collection of epidemiological time series focused on outbreak forecasting. IDOBE compiles from multiple data repositories spanning over a century of surveillance and across U.S. states and global locations. We perform derivative-based segmentation to generate over 10,000 outbreaks covering multiple outcomes such as cases and hospitalizations for 13 diseases. We consider a variety of information-theoretic and distributional measures to quantify the epidemiological diversity of the dataset. Finally, we perform multi-horizon short-term forecasting (1- to 4-week-ahead) through the progression of the outbreak using 11 baseline models and report on their performance. In addition to standard metrics such as NMSE and MAPE for point forecasts, we include probabilistic scoring rules such as Normalized Weighted Interval Score (NWIS) to quantify the performance. We find that MLP-based methods have the most robust performance, with statistical methods having a slight edge during the pre-peak phase. IDOBE dataset along with baselines are released publicly on https://github.com/NSSAC/IDOBE to enable standardized, reproducible benchmarking of outbreak forecasting methods.

A central question in computational neuroscience is whether the learning rule used to train a neural network determines how well its internal representations align with those of the human visual cortex. We present a systematic comparison of four learning rules (backpropagation (BP), feedback alignment (FA), predictive coding (PC), and spike-timing-dependent plasticity (STDP)) applied to identical convolutional architectures and evaluated against human fMRI data from the THINGS-fMRI dataset (720 stimuli, 3 subjects) using Representational Similarity Analysis (RSA). All models process stimuli at 224 x 224 resolution; results are averaged across 5 random seeds. Crucially, we include an untrained random-weights baseline that reveals the dominant role of architecture. At V1/V2, the untrained baseline exceeds backpropagation (rho = 0.076 vs. rho = 0.034; Delta-rho = +0.044, p < 0.001), and STDP achieves the highest V1 alignment among trained rules (rho = 0.064). At LOC, only BP reliably exceeds the random baseline (rho = 0.012 vs. -0.005, p < 0.001). At IT, all five conditions converge (rho = 0.008-0.014) with no significant pairwise differences among trained rules (p > 0.05, FDR-corrected). FA consistently produces the lowest alignment at V1, V2, and LOC (rho = 0.012 at V1, below all other conditions). Partial RSA confirms all effects survive pixel-similarity control. Seed variability is small relative to between-rule differences at V1/V2. These results demonstrate that early visual alignment is architecture-driven, learning rules differentiate only at intermediate areas, and all rules converge at the highest levels of the hierarchy.

Transfer RNAs (tRNAs) are essential components of the translational machinery. Their abundance and diversity shape decoding capacity and protein synthesis efficiency and accuracy. Because tRNA abundance is encoded in the genome through tDNA copy number, chromosomal organization, and cis-regulatory sequences controlling transcription, these features are expected to influence translational. However, the principles governing nuclear tDNA organization remain poorly understood. Here, we analyzed nuclear tDNA repertoires across 53 photosynthetic eukaryotes spanning major Archaeplastida lineages and secondary endosymbionts, along with seven non-plant eukaryotic outgroups, using comparative genomic approaches at sequence, chromosomal, and genome-wide scales. To support these analyses and enable interactive exploration of tDNA organization, we developed ShinytRNA (https://nebula.ibmp.unistra.fr/shinytRNA/), a web application for genome-wide exploration of tDNA organization. Nuclear tDNA copy numbers vary by more than two orders of magnitude across species, yet the relative representation of tRNA families corresponding to each amino acid remains strikingly conserved across lineages, revealing strong evolutionary constraints on tDNA dosage. In angiosperms, tDNAs show reinforced cis-regulatory features linked to RNA polymerase III transcription, including expanded AT-rich regions, enriched CAA motifs, and extended poly(T) tracts. At the chromosomal scale, tDNAs are predominantly dispersed along chromosome arms, with homogeneous spacing that scales with genome size, while also showing non-random chromosomal distribution, exclusion from centromeric regions, and limited clustering. Together, these patterns reveal conserved yet lineage-specific principles governing nuclear tDNA organization in plants, and highlight how multiple genomic constraints shape the evolution of nuclear tDNA repertoires.

Predicting species persistence within ecological communities is a fundamental challenge for both empirical and theoretical ecology. Existing methods span from mechanistic models, whose parameters are difficult to estimate from data, to statistical tools whose context-specific parameters are less interpretable. Here, we present a general framework, grounded in the statistical physics of complex systems, that integrates the key processes governing species survival into a single measurable quantity: the competitive balance. This metric quantifies a focal species' vulnerability beyond its abundance by incorporating the diversity of dispersal strategies and the structure of interspecific interactions within the community. Crucially, it can be inferred from spatial abundance data, thus circumventing the need to estimate species traits or dispersal parameters. Our results reveal that greater heterogeneity in dispersal strategies reduces vulnerability for a given abundance. Although we validate the framework using tropical and temperate forest data, it can be applied to a range of different ecosystems, providing a systemic and interpretable tool for assessing a context-dependent species vulnerability that accounts for its interactions with the entire community.

Accurate quantification of the geometry of curvilinear biological structures is essential for understanding cellular mechanics and disease-related morphological alterations. Microtubule curvature is a key descriptor of filament rigidity and mechanical perturbations. However, reliable curvature extraction from fluorescence microscopy images remains challenging due to noise, low contrast, and partial filament visibility. Existing approaches rely on segmentation pipelines with pre or post-processing, which are highly sensitive to segmentation errors and often fail under adverse imaging conditions. In this work, we propose MTCurv, a deep learning framework for direct, segmenta-tion-free regression of microtubule curvature maps from noisy microscopy images. Leveraging a synthetic dataset with pixel-wise curvature annotations, we reformulated curvature estimation as a regression problem and adapted an attention-based residual U-Net. To reduce hallucinations and enforce spatial coherence, we introduced a gradient-aware loss combining Mean Squared Error with a gradient consistency term. Beyond model and loss design, we evaluated commonly used regression and image quality metrics, revealing that many perceptual and blind metrics are poorly suited for curvature estimation. Correlation-based metrics, particularly Spearman correlation, emerged as more reliable indicators of curvature prediction quality. Experiments on two datasets of increasing difficulty demonstrated that MTCurv accurately recovers local microtubule curvatures, even in the presence of background fluorescence. Ablation studies highlighted the contribution of both residual encoding and attention-based decoding. Overall, this work provides a practical tool for filament curvature analysis and methodological insights for geometry-aware regression in biomedical imaging. Datasets and code are made available.

Experiments show that isochoric (constant-volume) conditions enhance supercooling stability relative to isobaric (constant-pressure) conditions. Here, combining Helmholtz equilibrium thermodynamics with a first-order perturbation methodology, we derive an inequality governing nucleation stability under volumetric constraint. The derivation provides a general thermodynamic proof that for any substance undergoing phase transformation in which the solid is less dense than the liquid, the Helmholtz driving force for solidification in isochoric systems is smaller than the Gibbs driving force in isobaric systems. Since nucleation rates depend exponentially on the inverse square of the driving force, this provides a thermodynamic basis for the observed suppression of nucleation rates. While a full stochastic treatment is beyond the scope of this work, the reduction in driving force implies a weakening of the bias toward growth of pre-critical fluctuations, increasing their probability of thermal dissolution. The analysis yields a dimensionless isochoric stability number. This number is computable from bulk thermodynamic data alone and provides a geometry-independent criterion for comparing metastable liquid stability across materials and conditions.

The evolutionary origins of structural features in reconstructed gene-regulatory networks (GRNs) remain poorly understood, especially given the random aspects of gene expression. Here, we extend a classical model of GRN evolution to allow a single network to express a distribution of phenotypes through noisy developmental dynamics. Inspired by Hopfield networks, we introduce an alignment score that quantifies the cohesion of gene-gene interactions in the network to support a target stable phenotype. Overall, evolved populations optimized their fitness and reduced the length of their developmental paths. Increased noise levels promoted alignment, enriched coherent feedforward and positive feedback loops relative to non-evolved and noiseless controls, and buffered against mutational perturbations. Alignment provides intuitive interpretations because an increased number of appropriately signed gene-gene interactions is more redundant and thus more robust against developmental noise and mutations. Together, these results demonstrate that cell-to-cell variability exerts strong selective pressure, driving the evolution of aligned, robust, and motif-enriched GRN architectures.

Black Sigatoka disease (BSD), also known as black leaf streak disease, is an airborne fungal infection caused by \textit{Pseudocercospora fijiensis} that severely impacts global banana and plantain production. Its persistence and resistance to eradication make it one of the most challenging plant diseases to manage. In this paper, we propose a deterministic pathogen-host model to describe BSD dynamics. Due to dual transmission pathways (ascospores and conidia) and mate limitation in sexual reproduction, the model exhibits a backward bifurcation: a stable endemic equilibrium coexists with the disease-free equilibrium for certain parameter values in which the basic reproduction number, $\mathcal{R}_0$, is less than 1. This phenomenon explains why control strategies that solely reduce $\mathcal{R}_0$ below one may fail. For the backward bifurcation regime, we perform sensitivity analysis of the endemic equilibrium using normalized forward sensitivity indices, Latin Hypercube Sampling, and Partial Rank Correlation Coefficients. Results indicate that effective control must extend beyond $\mathcal{R}_0$ reduction and prioritize (1) limiting production of new susceptible leaves during high-risk periods and (2) developing and deploying disease-resistant plant varieties. To incorporate transmission variability, we also formulate a stochastic version of the model using the Stochastic Simulation Algorithm (SSA). Extensive numerical simulations compare stochastic realizations with deterministic predictions and quantify variability in disease dynamics. To identify the principal drivers of persistence and variability, we analyze the endemic equilibrium using Sobol's variance-based sensitivity method, which highlights the role of nonlinear parameter interactions in shaping variability.

Identifying genes associated with diseases is crucial to understanding disease mechanisms and developing therapies. However, identification of individual genes associated with a disease often needs to be supplemented with clustering analysis to understand the relationships between genes and identify gene modules beyond individual gene-level relationships. Gene co-expression networks are widely used as a graph theoretic approach to the clustering analysis of genes. In our work, we perform robust clustering analysis on RNA-Seq data of Age-related Macular Degeneration (AMD) patients and controls by generalizing one such framework, Multiscale Embedded Gene Co-Expression Network Analysis (MEGENA). We propose a carefully curated set of module quality evaluation metrics to choose appropriate statistical distance-based or information theoretic similarity measures over simple linear correlation to represent the similarities between genes. Furthermore, we design and implement a stability test to ensure the robustness of the detected hub genes in the presence of noise. Finally, we propose differential module eigengene analysis for a deeper understanding of upregulation and downregulation of each module with respect to the disease and control groups for a comprehensive understanding of the clustering analysis. Besides detecting robust hub genes and modules that are supported by prior findings, we also identify previously undiscovered hub genes that can potentially lead to further biomedical research into understanding the AMD disease mechanism and developing new treatments.

Mosquito vector competence is usually represented as a process in which once virus is detected in saliva, mosquitoes are assumed to remain infectious for life, implying an irreversible transition to the transmitting state. However, some experiments report declines in the proportion of transmitting mosquitoes at late times post-exposure, suggesting transmission capacity may not be permanent. To investigate this hypothesis, we extended a previously developed stochastic intra-vector viral dynamics model by introducing transmission states allowing either permanent cessation or temporary interruption of transmission. We fitted three competing models to data from 52 vector competence conditions covering chikungunya, dengue, Zika, West Nile, and Rift Valley fever viruses, using Approximate Bayesian Computation with Sequential Monte Carlo inference. Among the 10 experimental conditions showing decline in transmitter proportions, models allowing exit from the transmitting state provided a better fit in 7 cases, with clear improvement in 5. In these cases, allowing interruption of transmission increased posterior estimates of the proportion of mosquitoes that crossed all intra-mosquito barriers, whereas estimates of infected and disseminated state durations were largely unchanged. In cases where intermittent transmission was selected, its performance was similar to that of permanent cessation with non-transmitting periods lasting several days. These results indicate that the assumption of lifelong mosquito infectiousness does not always provide the best explanation for vector competence data and may lead to underestimation of the proportion of mosquitoes that become capable of transmission. Incorporating time-varying transmission competence into intra-vector models could improve interpretation of vector competence experiments and refine epidemiological representations of arbovirus transmission.

Cardiac digital twins hold great promise for personalized medicine, but they currently depend on complex constitutive models of tissue mechanics that are often over-parameterized for the clinical context. To address this, we introduce CHESRA (Cardiac Hyperelastic Evolutionary Symbolic Regression Algorithm), a physics-informed machine learning framework that automatically derives simple strain energy functions from multiple experimental data sources. Using a normalizing loss function, CHESRA identified two new functions with only three and four parameters, respectively. These functions achieve high data fitting accuracy in experimental scenarios while enabling more consistent parameter estimation than state-of-the-art approaches, both in tissue benchmarks and 3D simulations. By combining biophysical constraints with data-driven discovery, CHESRA demonstrates how physics-informed learning can generate accurate, personalizable models for advancing cardiac digital twins and clinical decision-making.

Gene drive alleles bias their own inheritance to offspring. They can fix in a wild-type population in spite of a fitness cost, and even lead to the eradication of the target population if the fitness cost is high. However, this outcome may be prevented or delayed if areas previously cleared by the drive are recolonised by wild-type individuals. Here, we investigate the conditions under which these stochastic wild-type recolonisation events are likely and when they are unlikely to occur in one spatial dimension. More precisely, we examine the conditions ensuring that the last individual carrying a wild-type allele is surrounded by a large enough number of drive homozygous individuals, resulting in a very low chance of wild-type recolonisation. To do so, we make a deterministic approximation of the distribution of drive alleles within the wave, and we split the distribution of wild-type alleles into a deterministic part and a stochastic part. Our analytical and numerical results suggest that the probability of wild-type recolonisation events increases with lower fitness of drive individuals and with smaller local carrying capacity. Numerical simulations show that these results extend to two spatial dimensions. The role of the migration rate however, is less clear but has a lower impact. We further demonstrate that, in the event of wild-type recolonization, the probability of subsequent drive reinvasion decreases with smaller values of the intrinsic growth rate of the population. Overall, our study paves the way for further analysis of wild-type recolonisation at the back of eradication traveling waves.

The ability to fly through openings in vegetation allows insects like bees to access otherwise unreachable food sources. The specific visual strategies employed by flying insects during aperture negotiation tasks remain unknown. In this study, we investigated the visual and geometric parameters of apertures that influence traversing honeybees. We recorded honeybees flying through apertures with varying shapes and sizes using high-speed cameras to examine their spatial distribution patterns and trajectories during passage. Our results reveal that the flight of bees was, on average, along the bilateral center of the edges of the aperture irrespective of the size. When apertures were smaller, bees tended to also fly closer to the vertical center. However, for larger apertures, they traversed at lower vertical positions (closer to the bottom edge). The behaviors suggest that honeybees modulate their flight trajectories in response to spatial constraints, adjusting trajectory relative to aperture dimensions. When entering at off-center horizontal positions, bees tended to access the vertical center of the aperture, indicating altitude selection influenced by the curvature of the edge below. This behavior suggests an acute awareness of the vertical and horizontal spatial constraints and a preference for maintaining a curvature-dependent altitude that optimizes safe passage. Our analysis reveals that honeybees modulate speed and altitude above the ventral edge passing beneath them, maintaining a ventral optic flow magnitude within a preferred range. This relationship suggests a control mechanism where bees rely on visual information in a narrow ventrally directed field to navigate safely through confined spaces.