Understanding the structure of the brain, and how it changes with time and disease, is a core goal of structural neuroimaging. Contemporary approaches to structural brain analysis are dominated by voxel-wise, mass-univariate methods such as voxel-based morphometry (VBM). However, these techniques require images to be normalized to a standard template, which can obscure subject-specific geometric features. Normalization to a common stereotactic space can also be problematic when comparing groups with substantial brain pathology, lesions, or other anatomical abnormalities. Here, we introduce two complementary pipelines based on persistent homology (PH), a tool from topological data analysis, to quantify multiscale geometric features of structural T1-weighted MRI scans. Pipeline 1 quantifies regional thinning by applying the Euclidean distance transform to tissue masks in a slice-wise manner. Pipeline 2 uses \(α\)-filtrations to measure structural similarity between pairs of scans, capturing sulcal widening and ventricular enlargement. Synthetic experiments with controlled induced lesions showed that Pipeline 1 is best suited to between-subject analyses, whereas Pipeline 2 is better suited to within-subject designs. Applied to real-world data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), Pipeline 1 separated Alzheimer's disease (AD) from cognitively normal (CN) participants using single-modality T1-weighted MRI without nonlinear registration (ROC-AUC = 0.895), with peak effects localized to medial temporal regions. Pipeline 2 captured disease-related longitudinal change, with follow-up scans remaining closest to their own baselines and AD subjects showing greater short-interval change than CN subjects. Together, these pipelines provide interpretable topological biomarkers for cross-sectional group comparisons and longitudinal tracking.

Melanoma is an aggressive form of skin cancer. Survival rates are excellent if it is detected early but fall markedly if it metastasises. A key step in early tumour progression is the formation of cell clusters, which can promote metastasis. However, the mechanisms driving cell clustering, and the role of phenotypic heterogeneity in the dynamics of these clusters, remain poorly understood. In this work, we propose a system of ordinary differential equations that models cluster formation dynamics within a coagulation-fragmentation-proliferation framework. Using Bayesian inference, we fit this model to in vitro time-lapse microscopy data from two melanoma phenotypes-proliferative and invasive-to uncover the predominant mechanisms driving cluster formation and how these differ between phenotypes. Additionally, we provide preliminary insights into how clustering behaviour in co-cultures contrasts with that observed in monocultures. The model quantifies phenotypic differences in clustering dynamics: invasive cells in monoculture exhibit nearly threefold higher coagulation rates than proliferative cells, whereas proliferative cells display slightly higher proliferation rates. These differences align with known gene expression profiles. When applied to co-culture data, the model predicts hybrid coagulation behaviour of the clusters influenced by both proliferative and invasive cells but dominated by the invasive cells, and an elevated proliferation rate, suggesting a mutually beneficial effect of phenotypic heterogeneity on cell proliferation.

Functional connectivity varies across individuals due to genetic and environmental factors, yet classical twin models typically confound non-shared environment with measurement error and are largely limited to resting-state analyses. We hypothesized that: i) explicitly modeling measurement error from repeated fMRI sessions enables more accurate application of classical twin models (ACE/ADE) to functional connectivity; ii) model applicability depends on scan-length and parcellation granularity; iii) genetic and environmental effects on functional connectomes show differentiated functional modules across conditions. We extended ACE/ADE models to include a repeated-scan derived error term by analyzing monozygotic and dizygotic twins from the Young-Adult Human Connectome Project dataset. Genetic and environment variance components were estimated for all functional couplings across resting-state and task conditions, integrated across conditions using a minimum-error criterion, and analyzed using multilayer community detection across resolution scales. Functional couplings segregated into distinct categories characterized by shared environmental, additive, dominant, or epistatic influences, with a substantial fraction not meeting twin-model assumptions. Integrating across conditions revealed hierarchical community structure in genetic and environmental components observed across community resolution scales. Incorporating measurement error into twin models improves interpretability and applicability at the functional connectome level, revealing that genetic and environmental influences are structured into coherent, multiscale brain networks.

Information theory is a powerful framework to capture aspects of dynamical systems with multiple degrees of freedom. Mathematically, the dynamics can be represented as a continuous curve $\mathcal{C}$ on a suitable hyperplane in flat space and the Fisher information provides the norm of an infinitesimal displacement along this curve. In many applications, however, we do not have direct access to $\mathcal{C}$. Instead, we have to reconstruct the latter from a time-series of measurements (obtained as samples of size $n$), which are represented by an ordered set of points $\widehat{\mathcal{C}}$ on the same hyperplane. In this work, we calculate the bias of the Fisher information for large $n$, which provides a quantitative estimation for how accurately the dynamics of a system can be reconstructed from a given set of sampled data. Based on this result, we show that a clustering of the degrees of freedom reduces the bias and thus improves the accuracy with which the new system can be described with the same data. Inspired by a recent proposal for such a clustering, we provide a quantitive assessment of the loss of information, which allows to estimate how much information about the dynamics of a system can reliably be extracted based on a given set of data. We illustrate our findings in the case of a simple compartmental model. Although the latter is inspired by epidemiology, the results of this work are applicable to very general dynamical models with multiple degrees of freedom.

Citizen science is transforming how cognitive scientists study the human mind, and video games are at the heart of this shift. By embedding experimental tasks into engaging, game-like experiences, researchers can reach large, diverse populations while collecting rich behavioral data outside the lab. In this review, we explore how citizen science video games bridge the gap between players and participants, turning entertainment into large-scale cognitive research. Drawing on recent projects such as Sea Hero Quest and The Music Lab, we outline the key benefits of this approach: scalability, ecological validity, and public engagement. We also examine the challenges of designing games that are scientifically rigorous, ethically sound, and meaningful for both researchers and players. Through professional game developer insights, we highlight what it takes to develop a successful citizen science video game for cognitive science, and why this approach is still rare in the literature.

RNA-seq count data are often affected by read-to-gene alignment ambiguity, especially in high-dimensional transcriptomics. This type of ambiguity can be conveniently expressed through granular counts, namely fuzzy-valued observations of latent discrete quantities. We study a class of fuzzy-reporting mechanisms and show that, when reporting exploits graded membership, ignorability fails generically, leading to a coarsening-not-at-random structure. A hierarchical model is then introduced as a tractable instance of this construction and illustrated using RNA-seq data.

As transistor dimensions continue to shrink, binary devices are rapidly approaching their fundamental limits in power density. In response, multi-valued systems have attracted significant attention due to their enhanced information density. Among these, the ternary system stands out as the most practical option, being the closest integer base to (e), which is considered optimal for information efficiency. Despite the intrinsic advantages of DNA nanomaterials, such as programmability, energy efficiency, and massive parallelism, their application in ternary logic remains largely unexplored, particularly in the realm of ternary addition circuits. This gap can be attributed to a fundamental challenge: ternary logic requires circuits capable of recognizing and processing a far larger set of input combinations than binary systems, a task that existing models and techniques often struggle to accomplish. In this work, we propose a novel architecture for a ternary full adder. Our design includes a competitive blocking (CB) circuit that enables the recognition and computation of all possible three-input ternary combinations. Coupled with a dynamic concentration adjustment (CA) strategy, this approach significantly enhances the number of trits that can be processed. Biochemical experiments demonstrate that the CB circuit successfully yields the correct output digits for a ternary full adder, achieving 17-trit ternary addition. To our knowledge, this work represents the first successful DNA-based ternary adder, establishing a new methodological foundation for DNA computing and highlighting its considerable potential for scalable digital information processing.

During the development of an organism, cells must coordinate and organize to generate the correct shape, structure, and spatial patterns of tissues and organs, a process known as morphogenesis. The morphogenesis of embryonic tissues is supported by multiple processes that induce the precise physical deformations required for tissues to ultimately form organs with complex geometries. Among the most active players shaping the morphogenetic path are fine-tuned changes in cell adhesion. We review here recent advances showing that changes on cell adhesion, a local, pair-wise property defined at the cell-cell contact level has important global consequences for embryonic tissue topology, being determinant in defining both the geometric and material properties of early embryo tissues.

Intrinsically disordered proteins (IDPs) represent crucial therapeutic targets due to their significant role in disease -- approximately 80\% of cancer-related proteins contain long disordered regions -- but their lack of stable secondary/tertiary structures makes them "undruggable". While recent computational advances, such as diffusion models, can design high-affinity IDP binders, translating these to practical drug discovery requires autonomous systems capable of reasoning across complex conformational ensembles and orchestrating diverse computational tools at scale.To address this challenge, we designed and implemented StructBioReasoner, a scalable multi-agent system for designing biologics that can be used to target IDPs. StructBioReasoner employs a novel tournament-based reasoning framework where specialized agents compete to generate and refine therapeutic hypotheses, naturally distributing computational load for efficient exploration of the vast design space. Agents integrate domain knowledge with access to literature synthesis, AI-structure prediction, molecular simulations, and stability analysis, coordinating their execution on HPC infrastructure via an extensible federated agentic middleware, Academy. We benchmark StructBioReasoner across Der f 21 and NMNAT-2 and demonstrate that over 50\% of 787 designed and validated candidates for Der f 21 outperformed the human-designed reference binders from literature, in terms of improved binding free energy. For the more challenging NMNAT-2 protein, we identified three binding modes from 97,066 binders, including the well-studied NMNAT2:p53 interface. Thus, StructBioReasoner lays the groundwork for agentic reasoning systems for IDP therapeutic discovery on Exascale platforms.

Motivation: Multi-omics integration can improve cancer subtyping, but modality informativeness and noise vary across cancer types and patients. Existing graph-based methods optimize modality weights jointly with the classification objective and therefore lack independent reliability estimates, so low-quality omics distort patient similarity graphs and amplify noise through message passing. Results: We propose CMGL, a two-stage framework that estimates per-sample modality reliability through evidential deep learning and uses the frozen confidence scores to guide cross-omics fusion and graph construction. On four MLOmics cancer-subtype tasks and the 32-class pan-cancer task, CMGL consistently improves over the strongest baseline, surpassing it by 4.03% in average accuracy on the four single-cancer tasks. Its representations recover the PAM50 intrinsic subtypes of breast invasive carcinoma (BRCA), and the BRCA-trained model transfers without fine-tuning to kidney renal clear cell carcinoma (KIRC), stratifying patients into prognostically distinct groups.

Periods of low dissolved oxygen concentration -- hypoxia and anoxia -- threaten the health of aquatic ecosystems and the services they provide.Hypoxia is strongly influenced by temperature, but the different sensitivities and response functions of oxygen removal and production processes to temperature are not regarded in most models. Here we present OxyPOM -- Oxygen and Particulate Organic Matter, a nuanced temperature-aware process-based biogeochemical model. OxyPOM incorporates nuanced temperature sensitivities for the key oxygen-related processes photosynthesis, re-aeration, respiration, mineralization, and nitrification. Further sensitive variables like optimal light intensity, winter grazing inhibition, and pathogenesis are also represented. Our model was tested in an idealized water column experiment, representing a typical estuarine seasonal low-oxygen environment. Differences between nuanced and uniform temperature sensitivities affect seasonal patterns of oxygen-related processes, resulting in under- or overestimation during different times of the year, particularly with higher differences in summer. While these changes may balance in the overall annual oxygen budget, uniform sensitivities underestimate particulate organic carbon production by up to a factor of four along the year and overestimate nutrient concentrations. This nuanced approach to temperature sensitivity allows us to explore and test new hypotheses related to climate warming and heatwaves, addressing the ecosystem changes demanded by climate change models.

We calculate the moments and response functions of a nonlinear random recurrent neural network in the large $N$ limit. Our approach does not require averaging over synaptic weights and gives the first nontrivial term in a $1/\sqrt{N}$ expansion of general intensive-order correlation functions, proving a recent conjecture by Shen and Hu as a special case. Our results provide an analytical link between synaptic connectivity, correlations in spontaneous activity, and the response of a network to small perturbations.

Comparative analyses of phylogenetic trees typically require identical taxon sets, however, in practice, trees often include distinct but overlapping taxa. Pruning non-shared leaves discards phylogenetic signal, whereas tree completion can preserve both taxa and branch-length information. This work introduces a polynomial-time algorithm for set-wide completion of phylogenetic trees with partial taxon overlap. The proposed method identifies and extracts maximal completion subtrees that frequently appear across the source trees and constructs a weighted majority-rule consensus. Branch lengths are scaled using rates derived from common leaves. Each consensus subtree is inserted at the position that minimizes the quadratic distance error measured against information from the source trees, with candidate positions restricted to the original branches of the target tree. We demonstrate that the algorithm runs in polynomial time and preserves distances among the original taxa, yielding a unique completion that is order-independent with respect to the processing order of target trees. An experimental evaluation on amphibians, mammals, sharks, and squamates shows that the proposed method consistently achieves the lowest distance to the subset reference trees across subsets among all methods, in both topology and branch lengths. An open-source Python implementation of the proposed algorithm and the biological datasets utilized in this study are publicly available at: https://github.com/tahiri-lab/overlap-treeset-completion/.

Developing robust and clinically reliable EEG biomarkers requires evaluation frameworks that explicitly address cross population generalization in multi site settings such as Parkinsons disease (PD) detection. Models trained under i.i.d. assumptions often capture population specific artifacts rather than disease relevant neural structure, leading to poor generalization across clinical cohorts. EEG further amplifies this challenge due to low signal to noise ratio and heterogeneous acquisition conditions. We propose a population aware evaluation framework to assess the robustness and clinical reliability of EEG biomarkers under distribution shift. Using an n gram expansion strategy, we enumerate all cross population train test configurations across five independent cohorts, resulting in 75 directional evaluations. A nested cross validation design with integrated channel selection ensures prospective biomarker identification without population leakage. Results show that cross population transfer is asymmetric and that both accuracy and biomarker stability improve with increasing training population diversity, achieving up to 94.1% accuracy on held out cohorts. A theoretical analysis based on mixture risk optimization and hypothesis space contraction explains these trends, showing that multi population training promotes population robust representations. This work establishes a principled framework for learning robust, generalizable, and clinically reliable EEG biomarkers for multi site biomedical applications.

We instruct an AI agent to construct two separate agentic AI platforms: one for autonomous training of predictive ML models for human-human and virus-human PPI, and the other for inducing explicit general rules governing human-human and virus-human PPI. The first agentic AI platform for autonomous training of predictive ML models for PPI is designed to consist of five AI agents that handle autonomous data collection, data verification, feature embedding, model design, and training and validation on three-way protein-disjoint cross-fold datasets. For human-human and human-virus PPIs, the final three-way protein-disjoint ensemble achieves an accuracy of 87.3% and 86.5%, respectively. For cross-checking and interpretability purposes, the second agentic AI platform is designed to replace ML predictions with human-readable rules derived from protein embeddings, physicochemical autocovariance descriptors, compartment annotations, pathway-domain overlap, and graph contexts. For human-human PPI, it is defined by a two-rule induction, whereas human-virus is induced by a more complex set of weighted rules. The rules induced by the second agentic platform align with the SHAP-identified features from the predictive ML models built by the first agentic platform. Taken together, our work demonstrates the agentic AI's ability to orchestrate from data planning to execution, and from rule induction to explanation in ML, opening the door to various applications.

Large-scale neuroscience is generating rich datasets across animals, brain areas and behavioral contexts, yet our modeling efforts remains fragmented across isolated experiments. We argue that understanding behavior requires integrative neurocybernetic models: understandable dynamical models that capture the closed-loop coupling of brain, body and environment, treat the brain as a controller pursuing latent objectives, represent structured variation across scales, and scale to heterogeneous datasets. Such models shift the goal from predicting neural recordings in isolation to inferring the organizing principles that govern neural and behavioral dynamics. We outline a practical route toward this goal by combining nonlinear state-space models and meta-dynamical extensions with scalable inference, knowledge distillation, mixed open- and closed-loop training, and connectomics-informed architectures. By pooling complementary constraints from recordings, behavior, perturbations and anatomy, integrative neurocybernetic models can provide statistical amplification, few-shot generalization, and mechanistic insight into shared dynamical structure, individual variation, and the control objectives that govern behavior. This agenda offers a model-centric path from fragmented data to a mechanistic science of how brains produce behavior.

Microbiome studies increasingly indicate that disease-associated shifts cannot be understood from compositional changes alone. The functional architecture of microbial communities encoded in patterns of association among microbial gene families may reveal how these systems reorganize across biological conditions. Here, we present a network-based framework for characterizing microbiome rewiring across conditions. The approach combines condition-specific network inference, differential network analysis and pathway enrichment to identify interactions that are gained, lost or altered between groups, with a specific focus on sex-dependent differences. We apply the framework to inflammatory bowel disease, type 2 diabetes and atherosclerotic cardiovascular disease, comparing male and female specific microbial gene-family networks within each disease context. Across these settings, differential networks reveal extensive rewiring of microbial functional interactions, suggesting that microbiome alterations are shaped not only by changes in abundance but also by shifts in community organization. Importantly, pathway enrichment of rewired interactions uncovers functional signals that are not apparent from individual networks alone, highlighting latent disease and sex associated mechanisms. Code, data and supplementary information are available on the web site.

Tongue squamous cell carcinoma (TSCC) is an aggressive malignancy with marked biological heterogeneity and variable clinical outcomes. Although molecular profiling has improved understanding of TSCC heterogeneity, its clinical use remains constrained by invasive tissue sampling and limited representation of whole-tumor spatial complexity. Meanwhile, most radiomics studies in TSCC have focused on downstream clinical endpoints, and whether imaging can non-invasively reflect intrinsic molecular subtypes remains unclear. In this study, an integrated transcriptomic-radiomics framework was used to investigate the relationship between preoperative imaging phenotypes and molecular subtypes in TSCC. Transcriptomic data from 60 TSCC cases in The Cancer Genome Atlas were analyzed using unsupervised consensus clustering, followed by differential expression and functional enrichment analyses. Matched preoperative imaging data from The Cancer Imaging Archive were manually annotated for primary tumor regions, and radiomic features were extracted using PyRadiomics; group differences were assessed with the U-test. Two stable molecular subtypes, C1 and C2, were identified. Their biological differences were mainly associated with squamous epithelial differentiation, inflammatory signaling, and lipid metabolism, with C2 showing greater enrichment of immune-related pathways. In addition, 10 radiomic features differed significantly between the two subtypes, mainly wavelet-derived texture features from gray-level size zone, dependence, co-occurrence, and run length matrices (P=0.00202-0.0162). These findings support the potential of radiomics as a non-invasive approach for characterizing molecular heterogeneity in TSCC and provide an initial radiogenomic framework for biologically informed preoperative assessment.

The relationship between brain lateralization and cognitive functions is well-documented. The left hemisphere primarily handles tasks such as language and arithmetic, while the right hemisphere is involved in creative activities like drawing and music perception. Eye-tracking technology has shown the potential to reveal cognitive states by measuring ocular metrics such as pupil diameter and fixation duration. However, the ability to distinguish lateralized brain activity using these ocular metrics remains underexplored. Here, we demonstrate that pupil diameter and fixation duration can effectively classify left and right brain hemisphere activities. We obtained a considerably high classification performance, with an F1 score of 0.894. The results suggest that ocular metrics are robust indicators of lateralized brain activity and can be applied in cognitive monitoring and neurorehabilitation. Our future work expands on this by integrating these methods into real-time applications EyeBrain, potentially broadening their use across various cognitive and neurological domains.

Rare disease diagnosis increasingly relies on integrating genomic, phenotypic and transcriptomic evidence, yet these signals remain difficult to reconcile within a common interpretive framework. Here we present RareCollab, an LLM-powered framework for multimodal reasoning in Mendelian disease diagnosis that integrates more than 100 diagnostic evidence signals across DNA, RNA, phenotype, curated variant-level knowledge, and in-silico pathogenicity evidence. This design enables large language models to operate as calibrated, interpretable reasoning modules rather than as a single end-to-end ranker. We applied RareCollab to 890 patients from three cohorts, including 119 Undiagnosed Diseases Network probands with paired DNA and RNA data, constituting a large systematic benchmark for multimodal rare disease diagnosis under paired genomic and transcriptomic evaluation. In this real-world multimodal benchmark, RareCollab prioritized 94% of diagnostic genes within the top 10. Across recall thresholds from top 1 to top 10, it consistently outperformed proprietary phenotype-driven LLM baselines including Claude Sonnet 4.6 and GPT-5-mini by more than 25% on average and surpassed established state-of-the-art variant prioritization methods by 11%-24%. RareCollab also reshapes the diagnostic contribution of RNA evidence, which contributes to prioritization of the diagnostic gene in 35% of cases (42/119). Together, these results establish RareCollab as a scalable and interpretable framework for multimodal rare disease diagnosis.

Prenatal psychological stress affects 15-25% of pregnancies and increases risks of preterm birth, low birth weight, and adverse neurodevelopmental outcomes. Current screening relies on subjective questionnaires (PSS-10), limiting continuous monitoring. We developed deep learning models for stress detection from electrocardiography (ECG) using the FELICITy 1 cohort (151 pregnant women, 32-38 weeks gestation). A ResNet-34 encoder was pretrained via SimCLR contrastive learning on 40,692 ECG segments per subject. Multi-layer feature extraction enabled binary classification and continuous PSS prediction across maternal (mECG), fetal (fECG), and abdominal ECG (aECG). External validation used the FELICITy 2 RCT (28 subjects, different ECG device, yoga intervention vs. control). On FELICITy 1 (5-fold CV): mECG 98.6% accuracy (R2=0.88, MAE=1.90), fECG 99.8% (R2=0.95, MAE=1.19), aECG 95.5% (R2=0.75, MAE=2.80). External validation on FELICITy 2: mECG 77.3% accuracy (R2=0.62, MAE=3.54, AUC=0.826), aECG 63.6% (R2=0.29, AUC=0.705). Signal quality-based channel selection outperformed all-channel averaging (+12% R2 improvement). Mixed-effects models detected a significant intervention response (p=0.041). Self-supervised deep learning on pregnancy ECG enables accurate, objective stress assessment, with multi-layer feature extraction substantially outperforming single embedding approaches.

We study the dominant eigenstructure of positive-kernel Fredholm operators arising in multi-state structured population models, including integral projection models and age-structured McKendrick-type equations. To obtain a determinant-free and interpretable characterization of the leading eigenvalue and eigenfunctions, we introduce a reference-point operator, a rank-one construction at the kernel level that renders point evaluation well posed and induces a Markov-chain-inspired decomposition in the continuous-state setting. This yields convergent series representations of the stable distribution and reproductive value in terms of iterated kernels, together with an Euler-Lotka-type characteristic equation expressed through reference-point moments. The iterates admit a closed combinatorial form via ordinary partial Bell polynomials, providing an explicit bridge from transition kernels to genealogical quantities. Under a dominant spectral separation condition, satisfied for a broad class of kernels including Hilbert-Schmidt, Doeblin-type, and rank-one perturbations, the expansion converges at the spectral radius and organizes the leading eigensystem as a genealogical aggregation across generations. As applications, we derive demographic indicators-type reproduction numbers, generation intervals, and expected generation numbers-directly from continuous-state kernels, without discretization and without restrictive Hilbert-Schmidt assumptions. The resulting framework clarifies how ancestry-weighted initial-state information accumulates across generations to determine population growth and composition.

In temperate regions, respiratory virus epidemics recur on a yearly basis, primarily during the winter season. This is believed to be induced by seasonal forcing, where the rate at which the virus can be transmitted varies cyclically across the course of each year. Seasonal epidemics can place substantial burden upon the healthcare system, with large numbers of infections and hospitalisations occurring across a short time period. However, the interactions between seasonal forcing and the factors necessary for epidemic resurgence - such as waning immunity, antigenic variation or demography - remain poorly understood. In this manuscript, we examine how the dynamics of antibody waning and antigenic variation can shape the seasonal recurrence of epidemics. We develop a novel susceptible-infectious-susceptible (SIS) immuno-epidemiological model of respiratory virus spread, where the susceptible population is stratified by their antibody level against the currently circulating strain of the virus, with this decaying as both antibody waning and antigenic drift occur. In the absence of seasonal forcing, we demonstrate the existence of two Hopf bifurcations over the effective antibody decay rate, with associated periodic model solutions. When seasonal forcing is introduced, we identify complex interactions between the strength of forcing and the effective antibody decay rate, yielding myriad dynamics including multi-year periodicity, quasiperiodicity and chaos. The timing and magnitude of seasonal epidemics is highly sensitive to this interaction, with the distribution of infection timing (by time of year) varying substantially across the parameter space. Finally, we show that seasonal forcing can produce resonant damping resulting in a cumulative infection incidence that is less than would otherwise be observed.

Predicting cellular responses to genetic perturbations represents a fundamental challenge in systems biology, critical for advancing therapeutic discovery and virtual cell modeling. While large language models (LLMs) show promise for biological reasoning, their application to perturbation prediction remains underexplored due to challenges in adapting them to structured experimental data. We present SynthPert, a novel method that enhances LLM performance through supervised fine-tuning on synthetic reasoning traces generated by frontier models. Using the PerturbQA benchmark, we demonstrate that our approach not only achieves state-of-the-art performance but surpasses the capabilities of the frontier model that generated the training data. Our results reveal three key insights: (1) Synthetic reasoning traces effectively distill biological knowledge even when partially inaccurate, (2) This approach enables cross-cell-type generalization with 87% accuracy on unseen RPE1 cells, and (3) Performance gains persist despite using only 2% of quality-filtered training data. This work shows the effectiveness of synthetic reasoning distillation for enhancing domain-specific reasoning in LLMs.

Estimating single-cell responses across various perturbations facilitates the identification of key genes and enhances drug screening, significantly boosting experimental efficiency. However, single-cell sequencing is a destructive process, making it impossible to capture the same cell's phenotype before and after perturbation. Consequently, data collected under perturbed and unperturbed conditions are inherently unpaired, creating a critical yet unresolved problem in single-cell perturbation modeling. Moreover, the high dimensionality and sparsity of single-cell expression make direct modeling prone to focusing on zeros and neglecting meaningful patterns. To address these problems, we propose a new paradigm for single-cell perturbation modeling. Specifically, we leverage dual diffusion models to learn the control and perturbed distributions separately, and implicitly align them through a shared Gaussian latent space, without requiring explicit cell pairing. Furthermore, we introduce a sparsity masking strategy in which the mask model learns to predict zero-expressed genes, allowing the diffusion model to focus on capturing meaningful patterns among expressed genes and thereby preserving diversity in high-dimensional sparse data. We introduce \textbf{Doloris}, a generative framework that defines a new paradigm for modeling unpaired, high-dimensional, and sparse single-cell perturbation data. It leverages dual conditional diffusion models for separate learning of control and perturbed distributions, complemented by a sparsity masking strategy to enhance prediction of zero-valued genes. The results on publicly available datasets show that our model effectively captures the diversity of single-cell perturbations and achieves state-of-the-art performance. To facilitate reproducibility, we include the code in the supplementary materials.

Early endolysosomal and autophagic defects are among the earliest cellular alterations observed in Alzheimers disease (AD), yet the molecular drivers linking amyloid precursor protein (APP) metabolism to vesicle trafficking dysfunction remain incompletely understood. The APP-derived fragment C99 has emerged as a potential upstream mediator of intracellular toxicity, but its impact on organelle homeostasis and its modulation by metabolic interventions remain unclear. Here, we show that neuronal expression of human C99 in Drosophila induces profound vesicular abnormalities, impaired autophagic turnover, and disrupted mitochondrial quality control. Ultrastructural analysis revealed extensive accumulation of enlarged vesicular compartments, accompanied by reduced mitochondrial turnover and accumulation of aged mitochondria. Treatment with the ketone body beta-hydroxybutyrate (BHB) restored autophagic cargo clearance, improved mitochondrial turnover, and normalized vesicular ultrastructure. These protective effects required neuronal ketone transport, indicating a neuron-intrinsic metabolic mechanism. Proteomic mapping of the C99-associated interactome revealed that ketone treatment remodels networks enriched for vesicle trafficking and proteostasis pathways. Network prioritization identified the retromer component VPS35 as a candidate regulatory hub. Functional analyses demonstrated that depletion of VPS35 abolished the BHB-dependent restoration of autophagy, mitochondrial turnover, and vesicle morphology. Together, these findings suggest that ketone treatment restores mitochondrial quality control through a VPS35-dependent mechanism in C99 induced neurodegeneration, providing mechanistic insight into how metabolic interventions may restore intracellular homeostasis in Alzheimers disease.

Anti-viral therapies are typically designed to target only the current strains of a virus, a myopic response. However, therapy-induced selective pressures drive the emergence of new viral strains, against which the original myopic therapies are no longer effective. This evolutionary response presents an opportunity: our therapies could both defend against and actively influence viral evolution. This motivates our method ADIOS: Antibody Development vIa Opponent Shaping. ADIOS is a meta-learning framework where the process of antibody therapy design, the outer loop, accounts for the virus's adaptive response, the inner loop. With ADIOS, antibodies are not only robust against potential future variants, they also influence, i.e., shape, which future variants emerge. In line with the opponent shaping literature, we refer to our optimised antibodies as shapers. To demonstrate the value of ADIOS, we build a viral evolution simulator using the Absolut! framework, in which shapers successfully target both current and future viral variants, outperforming myopic antibodies. Furthermore, we show that shapers modify the distribution over viral evolutionary trajectories to result in weaker variants. We believe that our ADIOS paradigm will facilitate the discovery of long-lived vaccines and antibody therapies while also generalising to other domains. Specifically, domains such as antimicrobial resistance, cancer treatment, and others with evolutionarily adaptive opponents. Our code is available at https://github.com/olakalisz/adios.

The foundation of cognitive flexibility and higher-order intelligence lies in the functional structure and activity of brain networks, which can be dynamically configured by both external environments and internal states. However, decoding these dynamics from high-dimensional neural data remains a challenge. In this study, we propose a computational framework using Recurrent Neural Networks (RNNs) with neural dynamic constraints to model source-localized resting-state EEG data from $114$ participants. We aim to clarify the "triple brain network configurations" driven by exogenous and endogenous factors, including external stimuli, information processing tasks, and spontaneous activities. Our model identifies the parietal network as a critical hub supporting these multiple configuration patterns. Furthermore, we reveal that the anterior and posterior parietal regions exhibit distinct functional specializations under different stimulus modalities. By formalizing a triple configuration framework, this work separates latent factors of brain dynamics and underscores the computational significance of parietal regions in orchestrating higher-order intelligence.

Collective decision-making arises from individual agents integrating their own personal observations with information obtained from social partners. In many biological systems that exhibit collective decision-making, the process by which social information is produced, transmitted, and used is subject to two key constraints. First, individuals often do not observe the internal states or personal observations of their neighbors; instead, they observe neighbors' discrete actions. Second, agents often have limited attention, such that, at any given moment, only a subset of social partners influences decisions. Using methods from nonlinear dynamics, we show that either of these constraints can cause collective accuracy to become extremely sensitive to the weight individuals place on the information they receive from others. This sensitivity arises from the spontaneous formation of echo chamber-like states in which individuals receive and transmit homogeneous social messages. Under such conditions, collectives become locked in self-reinforcing states that prevent them from tracking changes in the environment. We reveal the mathematical basis of this phenomenon, and show that it emerges not only in generic models of collective decision-making but also in models developed to describe specific biological systems, including neural circuits, eusocial insect colonies, and mobile animal groups. Finally, we identify biologically plausible mechanisms through which individuals may reduce the risk of echo chamber formation and achieve robust yet sensitive collective decisions without requiring fine-tuning parameters. Our results reveal how fundamental constraints on communication shape the dynamics and reliability of collective decisions across diverse biological systems.

Compartmental epidemic models, grounded in mass-action kinetics, often assume homogeneous mixing. Although this neglects network structure, recent results show that for Poisson random graphs, the classical SIR model, especially the susceptible decay curve, matches the susceptible decay dynamics of its network counterpart. Motivated by this, we investigate whether the extended SIRI model with relapse from the recovered class admits a similar correspondence. SIRI dynamics arise in sevaral scenarios like spread of diseases with reactivation and behavioral contagion with relapse. We derive parameter relationships under which the pairwise SIRI model on a Poisson network closely follows the mass-action ODE trajectories. When transmission per contact is small relative to recovery, the susceptible and infectious trajectories of both systems align. This establishes conditions under which nonlinear SIRI dynamics on networks can be effectively approximated by tractable mean-field equations.