https://arxiv.org/api/HJm9equw5Fv1TZUmxl08jIFRf8w 2026-03-24T08:38:06Z 2425 15 15 http://arxiv.org/abs/2603.02952v1 2026-03-03T13:05:11Z

Background: Single-cell foundation models such as Geneformer and scGPT encode rich biological information, but whether this includes causal regulatory logic rather than statistical co-expression remains unclear. Sparse autoencoders (SAEs) can resolve superposition in neural networks by decomposing dense activations into interpretable features, yet they have not been systematically applied to biological foundation models. Results: We trained TopK SAEs on residual stream activations from all layers of Geneformer V2-316M (18 layers, d=1152) and scGPT whole-human (12 layers, d=512), producing atlases of 82525 and 24527 features, respectively. Both atlases confirm massive superposition, with 99.8 percent of features invisible to SVD. Systematic characterization reveals rich biological organization: 29 to 59 percent of features annotate to Gene Ontology, KEGG, Reactome, STRING, or TRRUST, with U-shaped layer profiles reflecting hierarchical abstraction. Features organize into co-activation modules (141 in Geneformer, 76 in scGPT), exhibit causal specificity (median 2.36x), and form cross-layer information highways (63 to 99.8 percent). When tested against genome-scale CRISPRi perturbation data, only 3 of 48 transcription factors (6.2 percent) show regulatory-target-specific feature responses. A multi-tissue control yields marginal improvement (10.4 percent, 5 of 48 TFs), establishing model representations as the bottleneck. Conclusions: These models have internalized organized biological knowledge, including pathway membership, protein interactions, functional modules, and hierarchical abstraction, yet they encode minimal causal regulatory logic. We release both feature atlases as interactive web platforms enabling exploration of more than 107000 features across 30 layers of two leading single-cell foundation models.

2026-03-03T13:05:11Z Ihor Kendiukhov http://arxiv.org/abs/2509.22997v3 2026-03-02T21:26:16Z

Cytotoxic T lymphocytes eliminate infected or malignant cells, safeguarding surrounding tissues. Although experimental and systems-immunology studies have cataloged many molecular and cellular actors involved in an immune response, the design principles governing how the speed and magnitude of T-cell responses emerge from cellular decision-making remain elusive. Here, we recast the T-cell response as a feedback-controlled program, wherein the rates of activation, proliferation, differentiation and death are regulated through antigenic, pro- and anti-inflammatory cues. By exploring a broad class of feedback-controller designs as potential immune programs, we demonstrate how the speed and magnitude of T-cell responses emerge from optimizing signal-feedback to protect against diverse infection settings. We recover an inherent trade-off: infection clearance at the cost of immunopathology. We show how this trade-off is encoded into the logic of T-cell responses by hierarchical sensitivity to different immune signals. Notably, we find that designs that balance harm from acute infections and autoimmunity produce immune responses consistent with experimentally observed patterns of T-cell effector expansion in mice. Extending our model to immune-based T-cell therapies for cancer tumors, we identify a trade-off between the affinity for tumor antigens ("quality") and the abundance ("quantity") of infused T-cells necessary for effective treatment. Finally, we show how therapeutic efficacy can be improved by targeted genetic perturbations to T-cells. Our findings offer a unified control-logic for cytotoxic T-cell responses and point to specific regulatory programs that can be engineered for more robust T-cell therapies.

2025-09-26T23:16:30Z 13 pages, 6 figures Obinna A. Ukogu Zachary Montague Grégoire Altan-Bonnet Armita Nourmohammad http://arxiv.org/abs/2603.00834v1 2026-02-28T23:04:29Z

Effective wound repair treatments rely on a clear picture of how cell proliferation and migration are coordinated during tissue restoration. Fibroblasts are key contributors to tissue restoration in the dermis, and modern imaging tools allow their cell-cycle progression to be observed directly, enabling comparison between experiments and computational models. Here we investigate how different stages of the cell cycle influence fibroblast-driven wound closure using the Discrete Laplacian Cell Mechanics (DLCM) framework driven by time-lapse microscopy data. \textit{In vitro} assays provide cell positions, migration behaviour, and cycle-stage information, and we show that incorporating proliferation, migration, and cell cycle arrest allows the computational model to reproduce the essential experimental trends. The results reveal that arrest in the G1 phase notably impacts the cell cycle dynamics and that the initial spatial arrangement of cycle states significantly affects wound closure. By linking single-cell cycle dynamics with emergent tissue behaviour this work establishes a quantitative approach for exploring how intracellular processes shape repair processes. More broadly, it demonstrates the value of integrating high-resolution data with cell-based mechanical models and provides a foundation for systematic \textit{in silico} evaluation of therapeutic interventions.

2026-02-28T23:04:29Z Erik Blom Qiyao Peng Leah Pomfret Richard Mort Stefan Engblom http://arxiv.org/abs/2603.00678v1 2026-02-28T14:42:50Z

The capacity to precisely edit genomes has outpaced our ability to predict the consequences. A cell can be genetically perfect and therapeutically useless: edited exactly as intended, yet unstable, drifting toward unintended fates, or selected for properties that compromise safety. This paradox reflects a deeper gap in how we evaluate biological intervention. Current frameworks excel at measuring what was done to a cell but remain blind to what the cell has become. We argue that this blindness stems from treating cells as collections of independent variables rather than as dynamical systems occupying positions on high-dimensional state manifolds. Drawing on Waddington's epigenetic landscape, we propose geometric stability as a missing axis of evaluation: the directional coherence of cellular responses to perturbation. This metric distinguishes interventions that guide cells coherently toward stable states from those that scatter them across the state manifold. Validation across diverse perturbation datasets reveals that geometric stability captures regulatory architecture invisible to conventional metrics, discriminating pleiotropic master regulators from lineage-specific factors without prior biological annotation. As precision medicine increasingly relies on cellular reprogramming, the question shifts from ``did the intervention occur?'' to ``is the resulting state stable?'' Geometric stability provides a framework for answering.

2026-02-28T14:42:50Z Prashant C. Raju http://arxiv.org/abs/2512.00306v2 2026-02-27T04:49:08Z

Virtual cell modeling aims to predict cellular responses to perturbations. Existing virtual cell models rely heavily on large-scale single-cell datasets, learning explicit mappings between gene expression and perturbations. Although recent models attempt to incorporate multi-source biological information, their generalization remains constrained by data quality, coverage, and batch effects. More critically, these models often function as black boxes, offering predictions without interpretability or consistency with biological principles, which undermines their credibility in scientific research. To address these challenges, we present VCWorld, a cell-level white-box simulator that integrates structured biological knowledge with the iterative reasoning capabilities of large language models to instantiate a biological world model. VCWorld operates in a data-efficient manner to reproduce perturbation-induced signaling cascades and generates interpretable, stepwise predictions alongside explicit mechanistic hypotheses. In drug perturbation benchmarks, VCWorld achieves state-of-the-art predictive performance, and the inferred mechanistic pathways are consistent with publicly available biological evidence.

2025-11-29T04:02:24Z Accepted at ICLR 2026 Zhijian Wei Runze Ma Zichen Wang Zhongmin Li Shuotong Song Shuangjia Zheng http://arxiv.org/abs/2503.01834v3 2026-02-24T17:44:55Z

Fibroblast cells play a key role in maintaining the extracellular matrix. During wound healing, fibroblasts differentiate into highly contractile myofibroblasts, which secrete extracellular matrix proteins like collagen to facilitate tissue repair. Under normal conditions, myofibroblasts undergo programmed cell death after healing to prevent excessive scar formation. However, in diseases like fibrosis, the myofibroblasts remain active even after the wound is closed, resulting in excessive collagen buildup and a stiff, fibrotic matrix. The reasons for the persistence of myofibroblasts in fibrosis are not well understood. Here, we show the existence of a mechanism where direct physical contact between a fibroblast and a myofibroblast is sufficient for fibroblasts to transition into myofibroblasts. We demonstrate that the fibroblast-myofibroblast transition can occur even in the absence of known biochemical cues, such as growth factor activation or mechanical cues from a stiff, fibrotic matrix. Furthermore, we demonstrate that contact-based fibroblast-myofibroblast activation can be inhibited by the Gαq/11/14 inhibitor FR900359, which prevents the formation of myofibroblasts. These findings provide new insights into the persistence of fibrosis despite therapeutic interventions, suggesting a potential strategy for targeting the fibroblast-to-myofibroblast transition in fibrotic conditions.

2025-03-03T18:55:41Z Vasuretha Chandar Benjamin M. Goykadosh Harikrishnan Parameswaran http://arxiv.org/abs/2602.19521v1 2026-02-23T05:22:21Z

Atherosclerotic plaques are fatty, cellular lesions that form in artery walls. The early plaque contains monocyte-derived macrophages, which are recruited to consume locally bound lipid deposits. Plaque progression is characterised by an imbalance in the rates of cell entry and exit from the plaque, which can occur if macrophages die in situ rather than leave by emigration. The mechanisms that regulate macrophage emigration are not well understood, but there is evidence that a chemotactic response can guide macrophages out of the plaque towards the artery wall lymphatics. In this paper, we develop a novel spatial model of the early plaque to study the implications of macrophage chemotactic emigration. Using mathematical analysis and numerical simulations, we investigate how the properties of the chemotactic response contribute to the spatial characteristics and lipid burden of the model plaque. Calculations of macrophage transit times are found to provide a reliable indicator of long-term plaque lipid burden, and also highlight the potential rate-limiting effect of the internal elastic lamina (IEL) on chemotactic emigration. When macrophage emigration is rate-limited by the IEL, we observe non-monotonic cell and lipid profiles that are associated with macrophage accumulation deep in the plaque. The model further predicts that when the chemoattractant penetrates only a short distance into the plaque, the proportion of emigrating macrophages may increase relative to that for a longer-range signal. The theoretical observations in this study can potentially be used to identify evidence of macrophage emigration in data from real atherosclerotic plaques.

2026-02-23T05:22:21Z Michael G. Watson http://arxiv.org/abs/2410.17420v3 2026-02-22T00:03:26Z

We propose a mathematical kinetic framework to investigate interactions between tumor cells and the immune system, focusing on the spatial dynamics of tumor progression and immune responses. We develop two kinetic models: one describes a conservative scenario where immune cells switch between active and passive states without proliferation, while the other incorporates immune cell proliferation and apoptosis. By considering specific assumptions about the microscopic processes, we derive macroscopic systems featuring linear diffusion, nonlinear cross-diffusion, and nonlinear self-diffusion. Our analysis provides insights into equilibrium configurations and stability, revealing clear correspondences among the macroscopic models derived from the same kinetic framework. Using dynamical systems theory, we examine the stability of equilibrium states and conduct numerical simulations to validate our findings. These results highlight the significance of spatial interactions in tumor-immune dynamics, paving the way for a structured exploration of therapeutic strategies and further investigations into immune responses in various pathological contexts.

2024-10-22T20:47:19Z 30 pages, 9 figures Chaos, Solitons & Fractals, 200: 116969 (2025) Martina Conte Romina Travaglini 10.1016/j.chaos.2025.116969 http://arxiv.org/abs/2404.16769v2 2026-02-21T23:58:44Z

Tumor cell migration within the microenvironment is a crucial aspect for cancer progression and, in this context, hypoxia has a significant role. An inadequate oxygen supply acts as an environmental stressor inducing migratory bias and phenotypic changes. In this paper, we propose a novel multi-scale mathematical model to analyze the pivotal role of Snail protein expression in the cellular responses to hypoxia. Starting from the description of single-cell dynamics driven by the Snail protein, we construct the corresponding kinetic transport equation that describes the evolution of the cell distribution. Subsequently, we employ proper scaling arguments to formally derive the equations for the statistical moments of the cell distribution, which govern the macroscopic tumor dynamics. Numerical simulations of the model are performed in various scenarios with biological relevance to provide insights into the role of the multiple tactic terms, the impact of Snail expression on cell proliferation, and the emergence of hypoxia-induced migration patterns. Moreover, quantitative comparison with experimental data shows the model's reliability in measuring the impact of Snail transcription on cell migratory potential. Through our findings, we shed light on the potential of our mathematical framework in advancing the understanding of the biological mechanisms driving tumor progression.

2024-04-25T17:23:29Z 29 pages, 8 figures Communication in Nonlinear Science and Numerical Simulations, 145: 108673 (2025) Giulia Chiari Martina Conte Marcello Delitala 10.1016/j.cnsns.2025.108673 http://arxiv.org/abs/2412.05191v2 2026-02-21T23:28:45Z

Go-or-grow approaches represent a specific class of mathematical models used to describe populations where individuals either migrate or reproduce, but not both simultaneously. These models have a wide range of applications in biology and medicine, chiefly among those the modeling of brain cancer spread. The analysis of go-or-grow models has inspired new mathematics, and it is the purpose of this review to highlight interesting and challenging mathematical properties of reaction--diffusion models of the go-or-grow type. We provide a detailed review of biological and medical applications before focusing on key results concerning solution existence and uniqueness, pattern formation, critical domain size problems, and traveling waves. We present new general results related to the critical domain size and traveling wave problems, and we connect these findings to the existing literature. Moreover, we demonstrate the high level of instability inherent in go-or-grow models. We argue that there is currently no accurate numerical solver for these models, and emphasize that special care must be taken when dealing with the "monster on a leash".

2024-12-06T17:15:57Z 42 pages, 7 figures J. Math. Biol. 91, 58 (2025) R. Thiessen M. Conte T. L. Stepien T. Hillen 10.1007/s00285-025-02243-8 http://arxiv.org/abs/2602.18909v1 2026-02-21T17:23:46Z

Cells often divide under mechanical confinement, where surrounding structures restrict shape changes during cytokinesis. Although forces generated during confined division have been measured experimentally, it remains unclear how confinement geometry and mechanics determine the transmitted force. Here we develop a minimal mechanical theory of cell division under confinement. Modeling the cell as an incompressible volume bounded by an interface with effective isotropic tension, we show that confinement restricts the set of mechanically admissible furrow shapes. As the furrow radius decreases, it reaches it reaches a confinement-induced minimum. Beyond this point, further ingression does not alter the interface shape, and both pressure and axial force saturate. We analyze force and pressure in rigid, soft, and strong three-dimensional confinement and demonstrate that a single geometric mechanism underlies these distinct cases. After rescaling force and length by the appropriate geometric scale, cells of different size and surface tension collapse onto a single universal curve. The relevant length scale is the cell size for rigid and soft confinement, and the confinement size in fully enclosing three-dimensional confinement. In soft confinement, environmental stiffness and spindle-generated axial forces determine the operating force and pressure, while the geometric constraint fixes the maximal attainable levels. In summary, our results show that mitotic force transmission and mitotic pressure during cytokinesis are bounded by confinement geometry, with material properties and active forces selecting the operating point within these geometry-imposed limits.

2026-02-21T17:23:46Z 7 pages, 3 figures Amit Singh Vishen http://arxiv.org/abs/2410.04512v2 2026-02-20T18:29:08Z

Off-lattice agent-based models (or cell-based models) of multicellular systems are increasingly used to create in-silico models of in-vitro and in-vivo experimental setups of cells and tissues, such as cancer spheroids, neural crest cell migration, and liver lobules. These applications, which simulate thousands to millions of cells, require robust and efficient numerical methods. At their core, these models necessitate the solution of a large friction-dominated equation of motion, resulting in a sparse, symmetric, and positive definite matrix equation. The conjugate gradient method is employed to solve this problem, but this requires a good preconditioner for optimal performance. In this study, we develop a graph-based preconditioning strategy that can be easily implemented in such agent-based models. Our approach centers on extending support graph preconditioners to block-structured matrices. We prove asymptotic bounds on the condition number of these preconditioned friction matrices. We then benchmark the conjugate gradient method with our support graph preconditioners and compare its performance to other common preconditioning strategies.

2024-10-06T15:05:18Z SIAM Journal on Numerical Analysis, 2026 Justin Steinman Andreas Buttenschön 10.1137/25M1727904 http://arxiv.org/abs/2507.12347v2 2026-02-18T15:19:27Z

The model organism Physarum polycephalum is known to perform decentralised problem solving despite absence of nervous system. Experimental evidence and modelling studies have linked these abilities, and in particular maze-solving, to some sort of memory and adaptation. However, despite compelling hypotheses, it is still not clear whether the tasks are solved optimally, and which key dynamical mechanisms enable Physarum's impressive abilities. Here, we employ a circuital network model for the foraging behaviour of Physarum polycephalum to prove that threshold sensing yields the emergence of unique and optimal paths that connect food sources and solve mazes. We also prove which conditions lead to alternative paths, thus elucidating how the organism achieves flexibility and adaptation in a self-organised manner. These findings are aligned with experimental evidences and provide insight into the evolution of primitive intelligence. Our results can also inspire the development of threshold-based algorithms for computing applications.

2025-07-16T15:40:13Z Daniele Proverbio Giulia Giordano http://arxiv.org/abs/2510.01935v2 2026-02-10T09:16:47Z

Background. Preeclampsia (PE) complicates 2-8% of pregnancies and involves placental hypoxia and HIF-pathway activation, especially in early-onset PE (eoPE). Chemical mimetics like cobalt (II) chloride (CoCl2) and oxyquinoline derivatives model trophoblast hypoxia in vitro, yet their fidelity in recapitulating PE gene profiles remains unclear. Integrating patient tissue analyses with experimental models may reveal common markers and validate physiologically relevant paradigms. Methods. We analyzed scRNA-seq data from 10 eoPE, 7 late-onset PE, and matched control placentas, identifying villous cytotrophoblast, syncytiotrophoblast, and extravillous trophoblast (EVT). BeWo b30 cells were treated for 24 h with CoCl2 (300 $μ$M) or the oxyquinoline derivative neuradapt (5 $μ$M) to induce hypoxia. RNA-seq with qPCR validation and small RNA-seq quantified mRNA and microRNA changes; PROGENy inferred pathway activities. Results. scRNA-seq revealed highest hypoxia activation in eoPE, with EVT showing maximum activity. Nine genes were upregulated across all trophoblast types (EBI3, CST6, FN1, RFK, COL17A1, LDHA, PKP2, RPS4Y1, RPS26). In vitro, neuradapt induced more specific hypoxia responses than CoCl2 (1284 vs. 3032 differentially expressed genes). Critically, EBI3, FN1, and COL17A1 showed concordant upregulation in tissue and neuradapt-treated cells, whereas CoCl2 produced opposite patterns. MicroRNAs hsa-miR-27a-5p and hsa-miR-193b-5p were consistently elevated in both models; 3'-isoforms of hsa-miR-9-5p and hsa-miR-92b-3p were identified as hypoxia-associated. Conclusions. EBI3, COL17A1, miR-27a-5p, and miR-193b-5p emerge as trophoblast hypoxia markers. Neuradapt (a selective HIF-prolyl hydroxylase inhibitor) provides a more physiologically relevant in vitro model than CoCl2, recapitulating transcriptomic signatures observed in PE placentas.

2025-10-02T11:57:21Z 31 pages, 5 figures, 1 table Placenta 176 (2026) 1-12 Evgeny Knyazev Faculty of Biology and Biotechnology, HSE University, Moscow, Russia Laboratory of Microfluidic Technologies for Biomedicine, Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry of the Russian Academy of Sciences, Moscow, Russia Timur Kulagin Faculty of Biology and Biotechnology, HSE University, Moscow, Russia Ivan Antipenko Faculty of Biology and Biotechnology, HSE University, Moscow, Russia Alexander Tonevitsky Faculty of Biology and Biotechnology, HSE University, Moscow, Russia Laboratory of Microfluidic Technologies for Biomedicine, Shemyakin-Ovchinnikov Institute of Bioorganic Chemistry of the Russian Academy of Sciences, Moscow, Russia 10.1016/j.placenta.2026.02.005 http://arxiv.org/abs/2602.10156v1 2026-02-10T00:57:38Z

Predicting how genetic perturbations change cellular state is a core problem for building controllable models of gene regulation. Perturbations targeting the same gene can produce different transcriptional responses depending on their genomic locus, including different transcription start sites and regulatory elements. Gene-level perturbation models collapse these distinct interventions into the same representation. We introduce STRAND, a generative model that predicts single-cell transcriptional responses by conditioning on regulatory DNA sequence. STRAND represents a perturbation by encoding the sequence at its genomic locus and uses this representation to parameterize a conditional transport process from control to perturbed cell states. Representing perturbations by sequence, rather than by a fixed set of gene identifiers, supports zero-shot inference at loci not seen during training and expands inference-time genomic coverage from ~1.5% for gene-level single-cell foundation models to ~95% of the genome. We evaluate STRAND on CRISPR perturbation datasets in K562, Jurkat, and RPE1 cells. STRAND improves discrimination scores by up to 33% in low-sample regimes, achieves the best average rank on unseen gene perturbation benchmarks, and improves transfer to novel cell lines by up to 0.14 in Pearson correlation. Ablations isolate the gains to sequence conditioning and transport, and case studies show that STRAND resolves functionally alternative transcription start sites missed by gene-level models.

2026-02-10T00:57:38Z 8 pages for main draft, 6 main figures Boyang Fu George Dasoulas Sameer Gabbita Xiang Lin Shanghua Gao Xiaorui Su Soumya Ghosh Marinka Zitnik