The adaptive immune system recognizes billions of unique antigens using highly variable T cell receptors (TCRs). To what extent TCR features shape the spatial distribution of the TCR repertoire (TCR-R) within lymphoid organs remains unclear. Here, we compared the TCR-R composition in T cell zones (TCZ), B cell zones (BCZ), and germinal centers (GC) of the spleen by analyzing the complementarity-determining region 3 of the beta chain (CDR3β) in microdissected splenic compartments from naïve and immunized mice. While we found substantial differences in TCR-R composition between these compartments, these differences largely disappeared once the effects of compartment size and proliferation were accounted for: During steady state, the few T cells found in BCZs strongly resemble random subsets of those found in TCZs. After immunization, expanded clones started appearing in the TCZ at day 3 and later shifted to the BCZ. GCs harbored even further expanded clones that were grouped into clusters with high sequence similarity. TCR-R differences during the immune response were well explained by a Galton-Watson model of T cell proliferation, suggesting that this could be the central factor shaping the observed TCR-R differences. Similarly, while TCR length and closeness to germline initially showed more pronounced differences between the compartments, our models of random distribution and proliferation explained these differences almost entirely. Hence, our findings suggest that splenic TCR-Rs are shaped predominantly by compartment size and proliferation rates rather than TCR sequence features. This would imply that quantitative assessment of TCR-R dynamics across heterogeneous spatial compartments during an immune response is only possible when these effects are explicitly accounted for.

Drug-drug interactions (DDIs) are critical in pharmacology, as synergistic or antagonistic effects between co-administered drugs can enhance therapeutic outcomes or cause toxicity, posing major challenges for drug safety assessment and new drug development. Current molecular representation methods often fail to capture dynamic substructure relationships or semantic interactions among functional groups. To overcome these limitations, this study proposes MEGH-DDI, a multimodal framework that integrates complementary molecular representation learning with graph-structure dynamic perception. The framework employs a dual-channel semantic module that combines Mol2vec-based substructure representations with MolT5 chemical language embeddings, while SHAP-based dynamic feature selection reduces redundancy. For fine-grained structural modeling, we introduce an edge-aware GCNII equipped with bond-level attention to distinguish different types of chemical bonds. Furthermore, the Hop2Token module converts multi-hop neighborhoods into token sequences, enabling Transformer-based modeling of high-order substructure dependencies. Experimental results demonstrate that MEGH-DDI achieves competitive performance across both transductive and inductive settings. On the DrugBank dataset, MEGH-DDI attains an accuracy of 98.36% in the transductive task, outperforming earlier baselines and surpassing recent state-of-the-art methods, with both AUC and AP exceeding 99.7%. On the ChChMiner dataset, MEGH-DDI achieves an F1 score of 99.05%, improving over the best existing method by 2.74 percentage points. In the inductive settings, MEGH-DDI achieves competitive performance across most evaluation metrics on both datasets, showing advantages over competing methods on several key metrics. These results demonstrate the effectiveness of MEGH-DDI in predicting DDI involving previously unseen drugs and its potential to generalize to complex DDI prediction scenarios. To enhance interpretability, MEGH-DDI incorporates a three-level mechanism spanning atoms, chemical bonds, and topological paths, highlighting key atoms and bond contributions. Overall, MEGH-DDI provides an efficient, accurate, and interpretable solution for drug safety assessment and new drug discovery. Code and datasets are available at https://github.com/Lilyjoys/MEGH-DDI, and a free web platform is provided at http://www.edgh-ddi.com.

Identifying cancer driver genes and their therapeutic impact remains a core challenge in computational cancer biology. We introduce xNNDriver and xAEDriver, two interpretable neural network frameworks that connect cancer mutations with genome-wide DepMap gene dependencies, pathway activity, and drug-response patterns. xNNDriver is a supervised pathway-guided model that evaluates whether a gene's mutation status is encoded in the genome-wide dependency landscape; we interpret model fitness as a driver potential score, which quantifies the strength of this mutation-dependency signal and prioritizes genes with broad functional footprints. Across 3,008 candidate genes, xNNDriver recovers major established drivers and highlights literature-supported candidates, while pathway analyses reveal biologically coherent programs related to metabolism, growth factor signaling, and immune regulation. To capture combinatorial functional states, xAEDriver uses an unsupervised autoencoder to learn Driver Variant Representations (DVRs), latent binary features guided by the frequency distribution of known driver mutations. DVRs capture cell-line-specific dependency patterns and expression patterns and are associated with drug sensitivity and pathway activity. Together, these interpretable deep learning models demonstrate that gene dependency landscapes encode rich, interpretable signals of oncogenic function and provide a hypothesis-generating framework for prioritizing drivers, pathways, and therapeutic vulnerabilities for further experimental validation.

Phase mapping is a widespread method for identifying rotational electrical activity sustaining cardiac arrhythmias. However, conventional implementations assume that the cardiac phase map is continuous, leading to ill-defined phase indices in regions affected by functional conduction block, fibrosis, or anatomical boundaries. These regions of discontinuous or undefined phase, termed phase defects, lead to both false positive and false negative detections of rotational drivers. This work introduces an improved phase mapping implementation termed extended phase mapping that explicitly detects and accounts for phase defects, enabling robust calculation of the phase index around them. Extended phase mapping is applied to (1) simulated excitation patterns using the Fenton-Karma model, (2) experimental optical mapping data of rat ventricular fibrillation, and (3) a clinical CARTO activation map of atrial tachycardia. Across all datasets, the extended approach eliminates erroneous detections and resolves previously missed rotations. Our results demonstrate that proper treatment of phase defects yields a unified and physiologically consistent characterization of all rotational drivers including near-complete and anatomical reentries. Therefore, we propose that explicitly accounting for phase defects, rather than tracking singular points alone, gives a more complete and consistent characterization of rotational dynamics in cardiac tissue.

Natural Killer (NK) cells are lymphocytes of the innate immunity and provide key protection against infection and cancer. NK cells express a wide array of inhibitory and activating receptors on the cell surface including the activating receptor NKG2D. Receptor signaling in NK cells leads to rapid post-translational modification (e.g., phosphorylation) of sub-cellular signaling proteins within minutes of receptor stimulation that ultimately give rise to diverse effector functions including cell proliferation. Recent single-cell mass cytometry (i.e., CyTOF) experiments in macrophages showed variations of abundances of phosphorylated signaling proteins across cell cycle stages indicating a dependence of cell signaling kinetics on an order of magnitude slower kinetics (~ several hours) of cell cycle transitions. We investigated cell cycle dependence of NKG2D signaling kinetics in NK cells using longitudinal CyTOF measurements up to 256 minutes post stimulation performed on IL-2-treated NKG2D-stimulated primary human CD56dim NK cells. The CyTOF experiments revealed cell stage specific kinetics of signaling proteins such as pVav1, pErk, and pS6, with the degranulation marker CD107a showing its largest upregulation specifically in the G2 + M stage. We developed stochastic minimal computational models coupling signaling and cell cycle processes to investigate the effect of cell cycle transitions on NK cell receptor induced signaling kinetics. These models suggest that the observed cell stage specific NK cell receptor signaling kinetics arise due to active modulation of the receptor signaling kinetics by cell cycle progression. The framework can be extended to study the interplay between receptor signaling and cell cycle progression in other cell types.

Understanding how structures arise in microbial communities remains an outstanding challenge in bacterial ecology. When replicating within dense and confined environments, bacterial populations may collectively self-organize as a result of mechanical interactions. We previously observed that model dual-populations of bacteria, colonizing open-ended microchannels, either quickly fixate to a single population or segregate into long-lived, coexisting populations. This latter quasi-stability results from the alignment of bacteria into lanes, forcing cells to grow toward the open ends of the channel, with inter-lane invasion events driving the boundary dynamics between populations. Here we apply agent-based (AB) simulations to explore the boundary dynamics between dual-bacterial populations of varying morphology and division rate. We find that the simulated boundary dynamics are well-described by a simple drift-diffusion model, which enables us to estimate the mean time to fixation within competitive scenarios where the fixation time is difficult to access by AB simulations. Coccus-like cells display a competitive advantage over bacillus cells as they can more effectively invade and ultimately fixate, while coexisting populations of bacillus cells are "effectively" stable. And while faster-dividing cells should have a selective advantage, their morphology determines if this advantage drives fixation or acts as a defensive strategy to maintain their population. These findings highlight the critical role that cell morphology and mechanics play in shaping bacterial communities.

Forecasting emerging infectious disease outbreaks is particularly challenging for sexual networks of men who have sex with men (MSM), where transmission can be shaped by heterogeneous and overdispersed contact structures, mixing assortativity, and both one-time and recurrent partnerships. Using stochastic epidemic simulations on synthetic MSM networks and empirical mpox case data from major cities, we systematically evaluated five classes of short-term forecasting models, from random walk approaches to semi-mechanistic attenuated renewal processes. Model performance was assessed using the continuous ranked probability score (CRPS). Across both simulated and empirical epidemics in heterogeneous networks, renewal models attenuated by a reverse-sigmoid curve, to account for early depletion of highly-connected nodes, consistently outperformed alternatives, achieving more balanced and reliable forecasts. Our findings suggest a practical approach for robust forecasting in highly overdispersed networks, where distinguishing network effects from other factors remains challenging and usually requires additional data.

Ecologists and conservation biologists rely on genetic diversity as a key essential biodiversity variable (EBV) used to track population health and dynamics, and utilize the population-scaled mutation parameter θ as a key metric of diversity. While whole-genome-sequencing (wgs) is increasingly affordable, it will be considerable time before the full diversity of life is represented by high-quality assembled genomes; even then, constant monitoring will still require repeated sampling across time and space. In contrast, genome skimming (low-coverage, short-read wgs) is highly cost-effective but challenging to analyze because the coverage is too low for assembly and reliable error correction. Mature methods, such as Mash, exist for estimating pairwise genomic distances based on the Jaccard similarity of k-mer sets computed using sketching techniques. Some, such as Skmer, additionally model the impacts of low coverage. These methods have been successfully applied to assembly-free species identification and phylogenetics; however, their use in population genetics has been limited. Among other limitations, these methods implicitly treat genomes as haploid, and heterozygosity confounds estimates of genomic distance for diploid organisms. In this paper, we address this problem through several technical advances. First, we use coalescent theory to mathematically derive how the Jaccard index between the k-mer sets of two diploid samples changes with θ under a neutral Kingman's coalescent model. Next, we model the impact of coverage and sequencing error to derive an estimate of θ from the Jaccard index, in addition to several auxiliary variables, which we also estimate from the genome skims. The resulting method, DipSkmer, enables more accurate estimates of coverage, sequencing error, and θ for diploid samples. Analyses of both simulated and empirical datasets show that for diploids and low distances (e.g., < 2%), DipSkmer's estimate outperforms existing alignment-free methods such as Mash and Skmer, and closely approximates ANGSD, a reference and alignment-based tool.

Glycoprotein VI (GPVI) and C-type lectin-like receptor 2 (CLEC-2) are platelet-activating receptors whose extracellular organization contributes to receptor association and signaling during thrombus formation, yet how alternative receptor architectures influence mechanical persistence remains unclear. We used molecular dynamics to compare a GPVI dimer based on 2GI7, a domain-swapped GPVI assembly derived from 7NMU, and a crystallographic CLEC-2 contact using the same pulling protocol and constant-force conditions. During the first 5 nm of imposed separation, reconstructed GPVI required less force and mechanical work than CLEC-2 and domain-swapped GPVI. Under constant-force loading from 400 to 600 pN, reconstructed GPVI also reached the same extension endpoint markedly earlier than the domain-swapped assembly, with passage times differing by more than an order of magnitude under several conditions. Extended pulling to approximately 10 nm produced complete loss of interreceptor heavy-atom contacts in reconstructed GPVI while local peptide and disulfide connectivity remained intact. In contrast, the domain-swapped preparation retained resistance over a larger part of the separation pathway and showed substantially greater mechanical persistence. These results show that alternative extracellular arrangements of the same receptor are not mechanically equivalent and that uncertainty in GPVI architecture can translate into large differences in force transmission and persistence under load. Such architecture-dependent persistence may influence the stability of extracellular receptor associations during thrombus formation and thereby affect the mechanical lifetime of receptor clusters under load.

Understanding non-genetic determinants of cell fate is critical for developing and improving cancer therapies, as genetically identical cells can exhibit divergent outcomes under the same treatment conditions. In this work, we present a deep learning approach for cell fate prediction from raw long-term live-cell recordings of cancer cell populations under chemotherapeutic treatment. Our Transformer model is trained to predict cell fate directly from raw image sequences, without relying on predefined morphological or molecular features. Beyond classification, we introduce a comprehensive explainability framework for interpreting the temporal and morphological cues guiding the model's predictions. We demonstrate that prediction of cell outcomes is possible based on the video only, our model achieves balanced accuracy of 0.94 and an F1-score of 0.93. Attention and masking experiments further indicate that the signal predictive of the cell fate is not uniquely located in the final frames of a cell trajectory, as reliable predictions are possible up to 8 h before the event. Our analysis reveals distinct temporal distribution of predictive information in the mitotic and apoptotic sequences, as well as the role of cell morphology and p53 signaling in determining cell outcomes. Together, these findings demonstrate that attention-based temporal models enable accurate cell fate prediction while providing biologically interpretable insights into non-genetic determinants of cellular decision-making.

In the geniculostriate pathway of the human visual system, neuronal projections carry signals from a particular retinal locus in parallel from one anatomical area to the next. Imprecision in the fidelity of these projections would place constraints on the ability of the system to perform tasks requiring positional information. We investigated the impact that "spatial scrambling" between stages would have on visual performance. We consider two stages in a simple canonical model of the early visual cortex where scrambling might occur: either the input to the first orientation-tuned mechanisms (analogous to V1 simple cells), or the output from those mechanisms. These are referred as "subcortical" (SCS) and "cortical scrambling" (CS). We developed a wavelet decomposition and resynthesis algorithm to mimic these effects, and measured human performance in letter identification affected by the two types of scrambling. Our results showed SCS and CS have distinguishable effects on both perceived noisiness of letters and letter identification threshold. Comparing human performance against a suite of pre-trained and custom convolutional neural networks (CNNs) that were trained on the scrambled stimuli, relative efficiency (calculated from the ratio of human:CNN thresholds) is higher for CS than SCS. However, in modelling human inefficiency by reducing the proportion of wavelets available to the CNNs, humans are less efficient in CS than SCS. These differences in efficiencies show humans are better at processing orientation redundant stimuli (CS) than orientation noisy stimuli (SCS). We hypothesize this reflects differences in integration properties at the input and output stages of simple cells in the cortex.

Neurons in the primary visual cortex (V1) are classically thought to encode spatial features of visual stimuli through simple population codes: each neuron exhibits a preferred orientation and preferred spatial frequency, both of which remaining invariant to other aspects of the visual stimulus. Here, we show that this simple rule does not apply to the representation of major features of stimulus motion, including stimulus direction and temporal frequency (TF). We collected an extensive dataset of cat V1 responses to stimuli covarying in orientation, direction, spatial frequency, and TF to assess the extent of motion selectivity. We show that preferred TF is mostly uniform across the cortical surface. Yet, in over half of V1, the preferred direction is reversed with changing stimulus TF, revealing four distinct map motifs embedded in V1's functional architecture. Similarly, despite the lack of spatial modulation for the preferred TF map and the lack of invariance for the preferred direction map, we found using convolutional neural networks that direction, TF and stimulus speed can be accurately decoded from V1 responses at all cortical locations. These findings suggest a latent encoding of information where subtle, spatially distributed modulations of V1 activity may convey fine information about stimulus motion, pointing to a novel primary sensory encoding mechanism despite complex co-variation of responses to multiple attributes across V1 neurons.

Antimony is a human-readable language for defining and sharing models developed by the systems biology community. It enables scientists to describe biochemical networks with a simple syntax, while supporting seamless conversion to and from the Systems Biology Markup Language (SBML) community standard. Since Antimony's original release, both SBML and modeling practices have evolved significantly, creating a need to update Antimony to maintain its standards compliance and practical relevance. In this paper, we introduce Antimony 3, a comprehensive update that formalizes its cumulative improvements and extends its support for SBML Level 3 Core and Flux Balance Constraints (FBC), Distributions, Layout, and Render packages. Antimony 3 enables model specifications that combine kinetic reactions with flux balance analysis, represent uncertainty using probability distributions, add biological context through annotations, and define publication-ready visualizations, all within a unified plain-text format. Antimony 3 is delivered as a lightweight C/C++ library with a stable C API. It is available through official bindings for Python, Julia, and JavaScript/WebAssembly, as well as a cross-platform desktop GUI, which enables straightforward use across scripting environments, desktop applications, and browser-based tools. Antimony 3 is released as open-source software under the BSD 3-Clause License and is available at https://github.com/sys-bio/antimony.

The complement component C4 regulates synaptic refinement and plasticity in the brain and has been implicated in multiple neurological and psychiatric disorders, with especially strong genetic evidence linking elevated C4A expression to schizophrenia. However, it remains unclear whether localized C4 elevation in a small neuronal population is sufficient to reshape broader cortical gene-regulatory programs and cellular phenotypes. To address this question, we overexpressed mouse C4 in approximately 2% of prefrontal cortical neurons using in utero electroporation and performed bulk RNA sequencing on microdissected prefrontal cortex. Despite the sparse manipulation, C4-OE was associated with broad transcriptional remodeling across tissue composed predominantly of untransfected cells, including coordinated changes in cholesterol biosynthesis, axon- and synaptic-associated programs, vascular/ECM pathways, and immune- and stress-related responses. Comparison with a schizophrenia synaptic proteomic dataset identified limited global correspondence but selective gene- and pathway-level overlap in specific molecular programs. Reference-based projection of directional DEG signatures onto a cortical single-cell transcriptomic reference further identified cell-type-associated structure, including prominent astrocyte- and vascular-associated axes and distinct lipid/metabolic, vascular/endothelial-interferon-associated, and immune/stress correlation domains. Spatial validation with MFISH showed increased Hmgcr expression in neighboring GFP-negative cells, while PLIN2 imaging revealed increased DAPI-associated/perinuclear lipid-droplet burden within the local C4-OE field. Together, these findings support a model in which sparse neuronal C4 elevation is associated with broader remodeling of the local cortical environment, including a non-cell-autonomous component involving neighboring-cell lipid-associated responses. These results expand the interpretation of C4 risk biology beyond synapse elimination alone and suggest that focal immune-gene dysregulation can engage lipid-homeostatic, stress-related, and cell-type-associated transcriptional programs relevant to cortical circuit vulnerability.

A key challenge in today's fast-paced digital world is to integrate information from various sources, which differ in their reliability. Yet, little is known about how explicit probabilistic information about the likelihood that a source provides correct information is used in decision-making. Here, we investigated how such explicit reliability markers are integrated into decisions and the extent to which individuals have metacognitive insight into this process. We developed a novel paradigm where participants viewed predictions from sources of varying explicit reliability to help them make a choice between two options. After each decision, they rated how much they felt a given source influenced their choice. Using computational modelling, we estimated the effective reliability that participants assigned to each source. Overall, we found that participants did not take source reliability at face value, inappropriately using the stated probabilities of source correctness to make a choice. Interestingly, even sources that were explicitly labelled as unreliable biased choices, as if these were treated as moderately reliable. Additionally, the presence of sources flagged as lying ones and known to be reliably predicting the incorrect answer impaired performance by increasing leakiness in evidence accumulation. Despite these biases, participants showed some metacognitive awareness of what influenced their choices: they were aware of the impact unreliable sources had on their decisions and could evaluate how much a given source had increased the likelihood of their response. These results suggest that people distort explicit source reliability but have some awareness of this process.