Papers

Single-cell trajectory inference maps transcriptomic measurements onto developmental continua, yet configurations that fit the data equally well can assign conflicting cell fates. FateMultiplicity is a label-free framework that constructs a statistically admissible model set, or Rashomon set, without lineage labels, by evaluating model discrepancy on cross-fitted held-out genes under non-inferiority testing calibrated against random-seed variation. Multiplicity is large and depends more on the diversity of the model space than its size: twelve configurations of a second algorithm expose 20.0% of cells where twenty-four of the first expose 3.8%. Whether the per-cell certified fate margin FM yields more reliable assignments than the fitted model already provides is then tested, and it does not. On simulation ground truth, on the same cells, FM discriminates misassignment at AUC 0.682, against 0.965 for the baseline configuration's own decision margin (p = 0.003) and 0.854 for a seed-dispersion baseline. Informativeness is governed by the breadth of the admitted set, not its cardinality: at cardinality four, seed refits give 0.933 and hyperparameter-perturbed sets 0.701. Relaxing the infimum to a q-quantile recovers discrimination but converges toward the single model's own confidence; the supremum reaches 0.973 because theta*'s membership bounds it from below, while the infimum is unanchored. Multiplicity in trajectory inference is worth measuring and reporting, but per-cell certification over a label-free Rashomon set is not a route to more reliable fate calls. Two constructions survive: a margin-erosion ratio separates real from spurious branch points in simulation (AUC 0.890, untested on real data), and against clonally observed fate, uncertified cells disagree with their clone's outcome 16.4 percentage points more often than certified cells (p < 0.001).

Theories of how the brain processes information and returns intelligent outputs are numerous and often difficult to conclusively test. Here, we consider an approach grounded in informatics and algorithmic thermodynamics for how neural systems respond to different patterns of information in different ways, with a focus on the relationship between the information entropy of a signal and a new proposed quantity we term information enthalpy, which represents the internal resource of structured information within a system available to be used for predictions and other forms of work. To evaluate the capacity of the external signals to increase information enthalpy in the system, we introduce the information enthalpy potential (IEP) as a defined metric. We propose and implement a method for quantifying the amount of potential information enthalpy - the IEP - in a given signal across a range of example signals. We offer a conjecture of how information enthalpy may be treated by neural systems within a biological framework before showing how it may integrate and offer falsifiability to existing approaches such as the Free Energy Principle. This framework is also positioned in light of the role of neural criticality. Finally, we postulate a testable and falsifiable framework where the distinct mechanisms within each driver interact through a described n-body-inspired model to govern the "motion" of a neural system through a high dimensional state-space. Through these processes, it is proposed that these features form the most fundamental basis of the neural drivers that give rise to the complexity of adaptive behaviors that are commonly called: intelligence.

The marked difference in reported human Lyme disease cases between the northeastern and southeastern United States remains a significant concern. Differences in biodiversity and the questing behavior of Ixodes scapularis, the primary vector of Borrelia burgdorferi, the causative agent of Lyme disease, have been proposed as key drivers of geographic variation in disease incidence. In this study, we investigate how these differences, together with ratio-dependent tick-host contact rates, contribute to the north-south gap in Lyme disease cases. Building on our previous work, we develop a discrete-time system of difference equations to model infection dynamics between ticks and their sylvatic hosts. We derive an expression for the basic reproductive number and vary key parameters to assess their impact on infection persistence. We further examine the effects of these mechanisms on nymphal infection prevalence (NIP) and the density of infected nymphs (DIN), two key metrics for predicting disease risk. Our results indicate that under baseline conditions, southern host communities exhibit lower infection prevalence and reduced potential for pathogen invasion, consistent with observed regional patterns. However, southern host communities can support sustained transmission with different questing behaviors. Notably, this qualitative north-south difference emerges using parameter values obtained from the literature, without calibrating the model to reproduce geographic differences in disease prevalence. In addition, ratio-dependent host-finding success shows a saturating effect on the basic reproductive number, compared to the traditional assumption of density-independent host-finding success. These results highlight the importance of biodiversity, tick questing behavior, and using multiple metrics when assessing Lyme disease risk.

Nonpharmaceutical interventions (NPIs), such as mask wearing, can reduce epidemic spread, but their effectiveness depends on whether people adopt and sustain them. The incentive to adhere changes with infection risk and with the behavior of others. Yet it remains unclear how epidemiological protection and social reinforcement determine the strategic structure of infection control, and how epidemic feedback changes that structure over time. Here, we develop a socio-epidemiological model in which prevalence changes behavioral incentives while adherence changes transmission, separating private protection from source control and incorporating conformity. At a given prevalence, private protection favors adherence, source control creates opportunities for free-riding, and conformity reinforces whichever behavior is already common, generating distinct adherence games. Because adherence reduces transmission and prevalence, the population can move between these games as the epidemic unfolds. This feedback can stabilize partial adherence, create alternative behavioral norms, or generate recurrent epidemic-behavior cycles. Moreover, interventions with the same population-level effect on transmission at a given adherence level can produce different long-term outcomes when protection is distributed differently between the individual and others. Thus, voluntary infection control is a feedback-evolving game whose outcome depends not only on epidemiological efficacy but also on who is protected and how behavior is socially reinforced.

In order to study oscillators coupled by pulses emitted at discrete times, it is generally assumed that the effect of each pulse is an instantaneous phase shift, with a possible conduction delay between emission and receipt. However, we are interested in neural oscillators in which each pulse (action potential) activates a biexponential synapse, and at high frequencies, the duration of the synaptic conductance may be greater than the network period. Therefore, the effect of received pulses can summate over the course of several cycles. To overcome this limitation, we previously defined a mean field approach for a single synchronous population. Here, we extend it to two alternating synchronous populations by assuming a perturbation of a single oscillator from one of the synchronous populations which is represented by a single self-connected neural oscillator. The train of delayed biexponential synapses emitted by each oscillator is divided into a tonic and a phasic component. Because of the continuous rather than pulsatile nature of the input, we had to drop the concept of oscillator phase and derive self-consistent criteria for the existence and stability of an alternating firing pattern between two synchronous populations based solely on the perturbation of time intervals from their steady state values. Simulations showed that for high frequency oscillations as defined above, the mean field approach predicts the existence and stability of phase-locking between two synchronous populations better than the instantaneous phase shift approach. This approach may generalize to other forms of coupling amongst non-neural oscillators.

Objective: To develop and validate a framework for spatially resolved ascending aortic strain estimation from 4D computed tomography angiography (CTA) and characterize reliability across segmentation sources and spatial scales. Methods: We constructed patient-specific aortic strain maps using nnU-Net segmentation, optimized surface remeshing, deformable registration, and mesh-based strain estimation. Finite element (FE)-derived synthetic 4D CT sequences provided controlled reference deformations for registration validation. Strain reliability was characterized across segmentation sources and Laplace-Beltrami (LB) spectral scales. Associations with diameter, aortic height index (AHI), and age were assessed. Results: Registration-derived median strain showed minimal bias relative to FE-reference strain for observer-derived (0.0013) and nnU-Net-derived meshes (0.0056). Median strain exhibited low bias and narrow 95% limits of agreement (LoA) across segmentation sources (interobserver: bias = -0.0045, LoA [-0.0490, 0.0399]; nnU-Net versus observer: bias = -0.0052, LoA [-0.0340, 0.0237]). Differences were greater for 95th-percentile strain, particularly in the interobserver comparison. LB spectral analysis showed preferential preservation of low-frequency strain organization, with k = 5 reconstructions reducing segmentation-related disagreement by 52-56%. The 95th-percentile areal strain was inversely associated with maximum ascending aortic diameter (r = -0.493, p = 0.005), AHI (r = -0.488, p = 0.005), and age (r = -0.479, p = 0.006). Three high-grade AR cases showed heterogeneous regional strain patterns. Conclusion: This framework reproducibly estimates ascending aortic strain from 4D CTA, characterizes its reliability across segmentation sources and spatial scales, and demonstrates its potential for noninvasive assessment of patient-specific aortic mechanics.

Molecular dynamics (MD) simulations model the motion of proteins between conformational states, but crossing free energy barriers and sampling multiple thermodynamic conditions remain computationally costly. We recently introduced MD-LLM-1, which showed that language models can generate conformational trajectories at lower cost, but required a separate model for each protein and did not explicitly model thermodynamics or kinetics. Here we introduce MD-LLM-2, a transferable language model that generates trajectories conditioned on sequence, temperature and molecular history. Because each step defines a normalized distribution over structural tokens, the model assigns an explicit likelihood to every generated or simulated trajectory. Trained across mdCATH domains, MD-LLM-2 reproduces the median radius of gyration of MD within 1 Angstrom in 10 of 12 unseen domains, predicts reduced Trp-cage compaction at higher temperature, and generates L99A T4 lysozyme trajectories that reach published transition-region backbone structures. In alanine dipeptide simulations, conditioning on a torsional scaling parameter $\lambda$ predicts basin populations at unseen couplings with a mean absolute error of 0.09 versus 0.35 for MBAR, and first-passage probabilities with an error of 0.08 versus 0.16 for a reversible Markov state model. Frame-isoentropic sampling further limits the accumulation of errors during autoregressive trajectory generation. Together, these results extend MD-LLM from protein-specific trajectory generation to transferable, physically conditioned modelling with explicit path likelihoods, and show that the framework can predict equilibrium populations for force-field parameters not seen during training, with an initial indication that first-passage kinetics can also be captured.

As AI models move toward clinical decision-making and personalized treatment, understanding what a model learns is important beyond predictive accuracy alone. We investigate whether personalized latent dynamics reveal clinically associated differences even when predictive fit is similar. A lightweight CNN–Transformer EEG foundation model pretrained on the Temple University EEG Corpus (TUEG) extracts segment-level representations. Using the Temple University Epilepsy Corpus (TUEP), representations are mapped to a shared latent-state space, and sparse multinomial logistic transition distributions (mLTD) are fit independently to each subject to obtain personalized transition-dependency graphs $W_n$. Analyses include $n{=}198$ subjects (99 epilepsy / 99 non-epilepsy). At $k{=}4$, epilepsy subjects exhibit substantially denser learned dependency structure ($p{=}1.1\times10^{-7}$), with the same pattern at $k{=}6$ (19.90 vs. 13.46; $p{=}5.2\times10^{-5}$). Graph-derived features provide moderate group discrimination under 5-fold subject-wise cross-validation (AUROC 0.68 at $k{=}4$; 0.65 at $k{=}6$). In contrast, held-out log-likelihood is nearly identical between groups at $k{=}4$ ($-0.992$ vs. $-0.991$; $p{=}0.95$), with similarly matched next-state prediction (AUROC 0.855 vs. 0.861; $p{=}0.54$). Thus, similar predictive fit does not imply similar learned dynamics: groups can be comparably predictable while differing substantially in the internal dynamical structure learned by personalized models. This distinction motivates evaluating learned structure alongside predictive performance in personalized clinical models.

Characterizing the genetic architecture of phenotypic variation is central to human genetics, yet trait polygenicity—the diversity of genetic variants responsible for a heritable trait's variation in a sample—is rarely quantified formally. O'Connor and Sella (2026) addressed this need by proposing four polygenicity measures that are quasi-arithmetic means and therefore satisfy a set of principles they argue are desirable for polygenicity measures. Here, I show that these four measures are points on a continuum corresponding to the Hill numbers, a family of ecological diversity statistics offering a tunable emphasis on variants with small versus large contributions. I further show that the Hill numbers are the only quasi-arithmetic means suitable for measuring polygenicity because they are the only ones that are "effective numbers" of variants. The connection to the Hill numbers clarifies the relationships among O'Connor and Sella's four polygenicity measures: all four are effective numbers of variants, and those that give greater weight to variants with small contributions yield higher estimates of polygenicity. It also connects polygenicity to results in the ecological diversity literature, including the replication principle, a criterion for how such measures should behave when independent sets of variants are combined. Together, these results offer a unified framework for interpreting, comparing, and selecting measures of trait polygenicity.

Recurrent neural networks (RNNs) offer linear-time scaling with sequence length while requiring only constant memory, yet they struggle to capture long-range dependencies due to vanishing gradients and limited receptive fields. To address these limitations, we introduce a second-order recurrent model in which the standard neuron-to-neuron communication is replaced by a spatially evolving field governed by (discretized) partial differential equations. Drawing inspiration from the role of cortical waves in brain computation, this mechanism allows structured spatiotemporal patterns to serve as an implicit, high-capacity memory. We show that the resulting model is equivalent to a structured infinite-order RNN in which the current state depends explicitly on its entire history of past states, yielding an effectively unbounded receptive field with a fixed number of parameters. We further derive constructive conditions to ensure marginal stability, constraining the gradient spectrum on the unit circle and thereby eliminating vanishing and exploding gradients. Empirically, the proposed architecture outperforms other recurrent models on long-horizon benchmarks while using substantially fewer parameters, demonstrating that spatial dynamics can effectively bridge the gap between efficient inference and long-term memory.

Taxonomic classifiers based on exact matches differ in which matches they use and in how each match splits its credit among the taxa containing it: Kraken~2 gives a $k$-mer's credit to the lowest common ancestor of those taxa, SPUMONI~2 labels each position of the read, Tagger gives each maximal exact match (MEM) one label, and Cliffy splits a match's credit evenly over a roughly logarithmic sample of the taxa. We compare these schemes on the same long MEMs, mainly on short 16S rRNA reads classified to genera with SILVA, where credit proportional to the number of occurrences of the MEM in each taxon is the most accurate, by 1.0 to 6.8 percentage points. We compute long MEMs with Boyer–Moore–Li on a run-length compressed suffix array, and count their occurrences in each taxon with a complete, run-length compressed tag array, in time proportional to the number of tag runs in the MEM's BWT interval, without locating any occurrence. We bound the number of tag runs in terms of the BWT runs and of how often suffixes with long common prefixes have different tags, and show that tags constant on $d$ blocks of the text have a macro scheme with at most $r + 2d$ phrases. On SILVA our index takes 1.44\,GB and classifies a read in 66\,$\mu$s with one thread at 93.8% accuracy, about what Cliffy reports for its 9\,GB index; over minimizer digests it takes 0.49\,GB and reaches 92.5%, against 84.2% for Cliffy's digested index on the same reads. On nanopore reads from whole genomes, where almost every long MEM occurs in one species, we instead use each read's longest MEMs, found in one pass of a variant of Boyer–Moore–Li with a $k$-mer filter, which assigns 99.6% of the reads to their species with no length threshold.

Quorum sensing (QS) is classically understood as a mechanism by which bacterial cells sense their own density, but environmental fluctuations make any single cell's estimate unreliable. An alternative view holds that populations pool individual estimates to reach reliable collective decisions. We investigate how a two-autoinducer toggle switch network enables this collective adaptation in fluctuating environments. At the single-cell level, the toggle switch is bistable. Membrane permeability controls autoinducer secretion and is the evolving trait. Using an individual-based model of a spatially structured population, we show that permeability evolves through a collective mechanism. Environmental frequency is the primary determinant of evolutionary outcome. Under strong asymmetry, the dominant trait reaches a high plateau. The trait linked to the rare environment stays low. It rises smoothly as its environment becomes more frequent. Total permeability stays close to a fixed budget; environmental frequency mainly sets how it is split between the two traits. When the two environments are equally frequent, no single trait wins. Both traits coexist, and noise asymmetry between the two sensing channels biases the levels at which they coexist rather than selecting a winner. Coexistence is also the slowest outcome to reach. Higher noise speeds adaptation, but the evolved permeability is non-monotonic in noise, peaking at an intermediate value. Earlier work found that more noise favors collective sensing as long as cells are correct on average. We develop a two-trait adaptive-dynamics model that unifies these regimes. Away from symmetry, it reduces to a closed one-dimensional equation that captures the individual-based results. At symmetry, weak cross-repression gives a single coexistence attractor, consistent with replicate populations maintaining both traits.

Single-cell RNA-sequencing analyses predominantly use either Seurat (R) or Scanpy (Python), with the choice often driven by programming language preference. However, by default, these tools produce divergent variable features, neighbour graphs, clusters, and marker genes. Laboratories requiring Python, which supports most deep-learning, foundation-model, and agent tools, must re-implement Seurat analyses in a framework that does not replicate Seurat's results. Truecell was developed as a Python implementation of the Seurat interface, preserving function names, arguments, default settings, and the object model to facilitate seamless transfer of analyses. In eighteen paired end-to-end evaluations against R Seurat, deterministic outputs matched to floating-point precision, and fold-change order was identical across all nine differential expression tests. In a three-arm benchmark on three datasets, with fixed user parameters and 20 seeds per tool, Truecell more closely reproduced Seurat's clustering than a Seurat-configured Scanpy in all 12 combinations of dataset and resolution settings. Truecell's marker genes and enriched pathways were also closer to Seurat's than Scanpy's were, and pseudobulk DESeq2 reproduced Seurat's gene lists with a Jaccard index ranging from 0.95 to 1.00. This agreement reflects fidelity to Seurat rather than biological correctness.

Cell viability assessment is a core requirement in cell culture systems, with critical applications in biopharmaceutical manufacturing and drug development. Conventionally, it is measured by adding membrane-impermeable dyes to a sample (a process called staining), which allows compromised cell membranes to be distinguished from intact ones. However, staining has several limitations: (a) chemical agents can perturb normal cellular processes of the cells being measured, (b) it is often ambiguous to assign viability to individual cells whose membrane integrity is only partially compromised. (c) photobleaching can undermine measurement accuracy over time when using fluorescent stains, and (d) staining cannot be performed in situ or in real time. Here, we show that (1) stained cells captured under brightfield imaging contain sufficient information to distinguish live and dead cells, and (2) cells captured under unstained brightfield imaging exhibit similar image features to their stained counterparts, enabling models trained on stained cells to generalize to unstained ones. We then report the development and validation of ViabiLens, an AI-assisted software for label-free cell viability analysis. The ViabiLens combines a cell detection model for localizing individual cells with a convolutional neural network (CNN) classifier for live/dead prediction, paired with an interactive UMAP-based viewer for visualizing and exploring individual cells across the sample. Evaluated on Chinese Hamster Ovary (CHO) cells spanning a wide range of viability conditions, ViabiLens achieves a mean absolute error of 2.68% on unstained samples against fluorescence-based reference measurements. We also release a benchmark dataset for label-free cell viability analysis to facilitate future research, available at https://amirrezavazifeh.github.io/ViabiLens-Project-Page/.

Understanding the extent to which measured synaptic wiring determines computation remains a central challenge. Here, we couple the proofread adult Drosophila melanogaster connectome to an anatomically faithful model of its eye. Visual information is inputted in the eye model, then passed to the connectome, and finally read from a Kenyon-cell-centered linear decoder. This creates a connectome-only model in which the anatomical graph and eye geometry are fixed and only scalar synaptic gains and neuronal thresholds may be learned. The model supports multitask vision, including color discrimination, shape classification, and numerical discrimination that follows a ratio-dependent scaling characteristic of approximate number perception. To test whether precise connectivity is consequential under wiring economy, we compare the biological graph to randomized ensembles that increasingly preserve biological synaptic constraints. At matched wiring cost, the biological network consistently yields higher accuracy, whereas less constrained rewiring surpasses it at the cost of inflated wiring. These findings indicate that the measured connectivity and eye geometry jointly set efficient operating points for visual computation.

Retrograde tracer injections in marmoset monkeys revealed a striking feature of neural circuitry in primate visual system: among neurons projecting to the primary visual cortex (striate cortex, V1), the proportion of extrastriate feedback neurons relative to thalamic feedforward neurons decreases markedly with visual field eccentricity. This supports the Central-Peripheral Dichotomy theory, which proposes fundamentally distinct computational specializations — visual recognition and attentional selection, respectively — in the central and peripheral visual fields.

Visual generative models are trained to capture the probability distributions of natural images, yet whether their native priors reflect the regularities governing human perception of image naturalness remains an open question. Here, we probe these priors through native prediction errors across 25 open image and video generators. Because raw single-image losses are dominated by scene content and visual complexity, we evaluate directional loss differences using content-preserving, paired relational interventions that selectively disrupt facial configurations or physical illumination consistency while limiting changes in low-level image statistics. Across both domains, these loss differences reproduce human-like selective sensitivities and tolerances, capturing the classic Thatcher effect on faces and shape-dependent responses to illumination inconsistencies. Notably, these loss differences reliably track continuous gradations of human naturalness judgments across individual stimulus pairs (peaking at $r = .84$ on faces and $.64$ on physical scenes) and retain unique human-aligned signals even after controlling for feature distances from frozen vision encoders and standard image quality metrics. We also find that while overall sensitivity to these violations broadly covaries with human alignment across models, the two systematically decouple along denoising schedules, with alignment peaking earlier than sensitivity, revealing that human-like naturalness judgments dissociate from generic violation detection. Together, these findings demonstrate that learning visual distributions yields generative loss landscapes that capture distinct aspects of human naturalness perception.

Using the recently published global census of the biomass of wild terrestrial mammals, we perform a detailed mathematical statistical analysis of its distribution over $N_s=4795$ species, drawing on tools developed in economics to characterize wealth inequality. We show that the Lorenz curve of the mammal biomass distribution is characterized by a large Gini coefficient $G=0.944$ exceeding the inequality reported for wealth distribution among world countries. This distribution is compared to the predictions of the Wealth Thermalization Hypothesis (WTH), in which species biomass values are treated as energy levels populated according to a Rayleigh-Jeans (RJ) steady-state distribution. We show that an RJ extended spectral model reproduces the real Lorenz and Pareto curves over four orders of magnitude of biomass variation, capturing the strong condensation of biomass among the rare, heavy-bodied species and its near-absence among the vast majority of light-bodied ones. These results extend the WTH framework, previously validated on distributions of human economic origin, to a biological distribution shaped by ecological and evolutionary constraints, and place the extreme rarity of large-bodied mammal species within the same statistical mathematical description as the oligarchic concentration of wealth in human societies. We finally discuss possible ecological mechanisms that may contribute to the observed distribution, including interspecific interactions, food availability, and competition.

Drug discovery for neurological disease has traditionally centered on the molecules altered by disease. But the molecules that cause pathology are not necessarily the best points from which to reverse it. Here, we ask which otherwise unaltered molecular control points can be engaged to restore pathological neural circuits toward functional states. We present CircuitATLAS, a provenance-grounded systems-neuroscience knowledge graph and agentic framework for target discovery in circuitopathies. It structures literature-derived relationships across diseases, phenotypes, electrophysiology, circuits, brain regions, cell types and molecular effectors, while deliberately excluding direct disease-gene and disease-protein edges to reduce shortcut reasoning. The graph contains 3.83 million nodes and 7.66 million edges, including 5.31 million LLM-extracted relations, and incorporates structured datasets such as the Human Cell Atlas and new multimodal in vivo measurements. We then introduce an agentic workflow that reasons from measurable disease phenotypes through their circuit and cellular substrates to molecular interventions, therapeutic feasibility and clinical constraints. Finally, we introduce a human-governed in vivo lab-in-the-loop linking hypothesis generation to experimental iteration. Within this framework an agent nominated ATP1A3, the neuronal alpha3 Na+/K+-ATPase, as a control point on cortical excitability; interneuron-restricted expression of ATP1A3 abolished the beta- and gamma-band response to a focal 4-aminopyridine challenge in vivo, and the validated target was then carried into a structure-guided small-molecule campaign terminating in a defined assay to resolve the direction of modulation. CircuitATLAS thus provides a framework for discovering therapeutics based not only on what is molecularly disrupted in disease, but on what can be controlled to restore circuit function.

Early cognitive response after repetitive transcranial magnetic stimulation (rTMS) may predict later cognitive change in Alzheimer disease, but the incremental value of domain-specific response is unclear. We analyzed data from 112 randomized participants; 101 entered a Week-5 landmark cohort and 91 had complete predictors and outcomes at 8, 16, and 24 weeks post-treatment. Nested predictor-set comparisons used ridge regression to evaluate baseline information, trial variables, early total-score response, and domain-specific response. Model-class comparisons included ridge, multi-task elastic net, restricted random forest, and longitudinal mixed effects. Validation used repeated treatment-site-stratified 5-fold cross-validation with five repetitions, nested hyperparameter tuning, and paired participant-level bootstrap comparisons with 10,000 resamples. Adding Week-3 and Week-5 ADAS-Cog total-score response improved prediction. The early total-response ridge achieved overall MAE 3.067, RMSE 3.885, and $R^2=0.384$. Replacing two total-response variables with 22 component-level changes worsened performance (MAE 3.363, RMSE 4.269, $R^2=0.257$). The paired MAE difference was -0.297 points (95% CI -0.485 to -0.115). Mixed effects was competitive (MAE 3.162), with an uncertain overall difference from total-response ridge. Sensitivity analyses using absolute future scores and excluding Week-3 response supported the main findings. In this sample, early aggregate ADAS-Cog response was more informative than a higher-dimensional domain-specific representation. The model is an internally validated early-response updating tool, not a pretreatment selection model or evidence of causal rTMS efficacy.

Familial amyloid-$\beta$ (A$\beta$) variants alter aggregation kinetics, but their interaction with structural brain connectivity remains incompletely understood. We developed a mutation-aware mechanistic model coupling a coarse-grained monomer–oligomer–fibril aggregation–fragmentation system to graph diffusion on the 540-node Budapest Reference Connectome component. Experimental A$\beta_{42}$ nucleation scores scaled primary nucleation rates for seven variants relative to wild type. Robustness was examined using mutation-score uncertainty, alternative kinetic mappings, global sensitivity analysis, seed and edge-weight perturbations, degree-preserving randomized connectomes, spatial propagation analysis, synthetic ABC-SMC parameter recovery, posterior prediction, and Chemical Langevin simulations. E22G showed the earliest threshold crossing and greatest cumulative oligomer burden, whereas A2V was delayed under the selected mapping. Mutation rankings persisted across tested network perturbations, although regional burden patterns depended on topology. Connectome distance from seed regions was associated with later oligomer arrival (Spearman $\rho\approx0.92$). Under inferred parameter uncertainty, the mean timing-rank correlation was 0.990 and cumulative-burden ordering was preserved in every posterior draw, while exact peak-amplitude ordering was less stable. Stochastic ensemble medians retained the deterministic ordering despite overlap among trajectories. Within this proof-of-concept framework, mutation-dependent kinetics primarily influence aggregation timing and cumulative burden, while connectivity shapes spatial propagation. Nominal background rates, model-time units, and synthetic parameter recovery limit interpretation to mechanistic comparisons rather than clinically calibrated prediction.

Recent advances in generative protein design have enabled de novo antibody generation with explicit target and epitope conditioning. However, most existing approaches remain formulated around a single antigen-antibody interface, whereas bispecific antibody design requires modeling multi-component complexes in which multiple target-recognition interfaces must coexist and interact within a shared antibody structure. Here we present OFAntibody, an all-atom generative framework for de novo design of monoclonal and bispecific antibodies. OFAntibody expands CDR-epitope interaction learning with large-scale distilled antigen-antibody complexes, and introduces multi-component structural supervision and arm-aware multi-hotspot routing to learn compatible multi-interface geometries and couple each antibody paratope to its designated epitope. OFAntibody supports epitope-conditioned generation across monoclonal antibodies and diverse bispecific formats, including tandem VHH, diabody and CODV. In nanobody design benchmarks, OFAntibody achieves a Top-5 enrichment rate of 41.5%, representing a 5.39-fold improvement over RFantibody in competitive candidate ranking. In bispecific antibody design tasks, OFAntibody achieves hotspot pass rates of 94-100% across the three evaluated tasks and achieves energy pass rates of 60%, 13% and 8% for diabody, tandem VHH and CODV formats, respectively. OFAntibody further enables the same target combination to be explored across different antibody formats, while joint multi-interface generation reduces geometric incompatibilities arising from independent design and post hoc assembly. Together, these results extend de novo antibody design from binary antigen-antibody complexes to programmable multi-component complexes, providing a foundation for designing single molecules that combine recognition of distinct targets and their associated biological functions.

Visual decoding from fMRI is typically siloed by participant and experiment, obscuring whether heterogeneous neural measurements can be organized within a common computational geometry. Here we introduce BRAID-fMRI (Brain Representation Alignment across Individuals and Datasets), a shared CLIP-supervised decoding framework. BRAID-fMRI uses a single ROI-wise Transformer with optional participant conditioning across eight visual-fMRI datasets comprising 93 dataset-specific participant entries, 430,007 single-trial responses and 162,839 unique stimuli. Regional brain activity is aligned with 512-dimensional CLIP ViT-B/32 representations using a multi-positive contrastive objective that treats repeated stimuli across participants and datasets as positives. BRAID-fMRI supports retrieval across seven evaluation datasets. On eight matched participant entries, it achieves 35.0 +/- 11.1% Top-10 accuracy, exceeding the observed mean accuracy of the two evaluated baselines - the MindEye-style pooled-CLIP decoder (27.1 +/- 5.1%) and ridge regression (21.1 +/- 9.1%) - and attaining the highest observed accuracy for seven of eight entries. In separately trained participant-agnostic models, expanding the source pool increased target-dataset holdout accuracy by up to 92.1% relative to the initial source-training condition. The learned space preserves graded semantic structure, while ablations and saliency highlight ventral and early visual cortex and category-specific motion and attentional systems. These results support scalable cross-dataset decoding into a common CLIP-aligned space, with model sensitivity concentrated in ventral and early visual inputs.

Naturalistic movie-fMRI provides a shared, temporally structured probe of brain dynamics, yet predictive models commonly rely on whole-run functional connectivity (FC) or temporally generic representations that are not aligned with narrative events. We introduce MovieSTAGE (Scene, Transition, and Global Encoding), a multiscale framework that combines hypergraph-structured FC-profile organization within scenes, unsigned FC-profile differences across adjacent scenes, and whole-movie FC. We evaluated 260 participants from the CMI-HBN Despicable Me cohort on case-control, ADHD-subtype, and three-class classification using 10 repetitions of stratified five-fold cross-validation, complete out-of-fold (OOF) predictions, and paired subject-cluster bootstrap and permutation tests. MovieSTAGE achieved AUROCs of 0.69, 0.73, and 0.75 and balanced accuracies of 67.6%, 69.8%, and 58.3%, respectively, yielding the highest mean point estimates among the evaluated methods. On the three-class task, the full model outperformed all two-branch variants, the HGNN scene encoder outperformed MLP, GAT, and BNT alternatives under matched settings, and the human-annotated partition outperformed duration-matched random and fixed-count GSBS controls. These controlled results support incremental predictive value from event-aligned scene and transition representations when combined with whole-movie FC in this cohort. Post-hoc model-derived analyses generated network-level hypotheses involving frontoparietal and default-mode systems.

Many rare genetic diseases are associated with recognizable craniofacial features. However, traditional approaches for describing facial morphology rely largely on qualitative clinical observation and free-text descriptions, which are often subjective, non-standardized, and difficult to reproduce across observers and institutions. Although the Human Phenotype Ontology (HPO) provides controlled terms for describing facial features, these terms are typically categorical rather than quantitative and may vary depending on examiner experience and interpretation. Here, we present FaceKit, a computational framework for quantitative facial phenotyping from frontal facial photographs. FaceKit extracts standardized measurements of facial landmarks and derived 120 morphological features, then reports feature-level z-scores representing deviation from population reference distributions. The reference distributions are built from the FairFace dataset spanning diverse ancestral groups. We evaluated FaceKit on a curated subset of the GestaltMatcher Database covering 50 rare-disease cohorts. In addition to quantitative facial analysis, FaceKit includes synthetic facial image generation to support rare disease model development and data augmentation. We also performed privacy evaluation to assess whether synthetic images reveal identifiable information from real patient photographs and could compromise patient privacy. Across disease case studies, FaceKit-derived quantitative measurements captured known facial features associated with rare genetic disorders and provided objective support for clinical phenotyping. Together, these results establish FaceKit as a useful tool for quantitative phenotyping, and has the potential to improve rare disease diagnosis, support genotype-phenotype studies, and enable more reproducible clinical characterization across diverse patient populations.

Naturalistic visual decoding from EEG has long been constrained by cohort size: existing scaling literature caps out at a few dozen subjects, leading prior work to conclude that scaling subject cohorts yields minimal performance gains. Training a cross-modal EEG–video contrastive encoder on a cohort larger by more than an order of magnitude, we find the axis is productive but has an onset. Below $S\approx50$ — the entirety of the range prior work occupies — no model improves meaningfully over an untrained encoder; above it, decoding rises log-linearly in subject count with no saturation at the top of our ladder, on a fitted stimulus-feature probe and on fit-free movie-moment retrieval alike, and the gain transfers to a second, unseen film. How far the axis carries then depends on initialisation more than on capacity: an encoder initialised from an EEG foundation model scales ${\sim}1.6\times$ faster per doubling of the cohort than randomly initialised encoders at two depths, is the only one still converting subjects into retrieval accuracy at the top of the ladder, and converges on a fraction of the alignment compute. Subject scaling on shared naturalistic stimuli is thus an effective and currently unsaturated frontier for scaling EEG foundation models.

Dasgupta, Stevens and Navlakha (2017) showed that the Drosophila olfactory circuit, modelled as a random sparse projection followed by winner-take-all, is a locality-sensitive hash that beats classical LSH. The projection was random because the wiring was unknown. We test it against four electron-microscopy connectomes (MaleCNS, hemibrain, FlyWire, BANC; four animals, seven hemispheres). First, the 2017 pattern holds in a reimplementation of its protocol on SIFT, MNIST and odour mixtures (GloVe is near chance at short codes for every method): the fly hash beats k Gaussian projections at short hash lengths (3.1x in AP@200 on MNIST at k = 4). Second, against the tested real-valued Gaussian baseline that advantage is per active cell, not per operation: Gaussian projections given the same projection arithmetic retrieve better on every dataset and input dimension tested. Third, across four connectomes the measured pairing of glomeruli gives no consistent retrieval advantage over degree-preserving rewiring: retrieval is slightly lower (median -1.6%), and the small odour deficits depend on how missing odour responses are treated. Separately, equalising glomerular fan-out at fixed connection count improves retrieval in the model in every hemisphere, while equalising inputs per cell lowers it on average. Yet the fan-out profile is similar across the four sampled animals (median between-animal Spearman rho = 0.86), structural synapse counts do not offset it, and its relation to odour tuning is weak. In the model that uneven allocation costs retrieval. For practice: measured wiring gives no consistent retrieval advantage over degree-preserving random wiring, so a fly hash needs no connectome data, and its advantage is per active unit, which may suit hardware where active units rather than arithmetic are the binding cost, a hypothesis we do not test.

Predicting transcriptional responses to genetic perturbations is central to understanding gene function. Existing predictors primarily rely on transcriptomic measurements, although chromatin accessibility provides complementary information about the cellular context in which perturbations act. Using this information requires linking chromatin context to specific perturbations and accounting for baseline differences between independently sampled control and perturbed populations. We propose ChromaPert, a chromatin-informed framework for predicting state-dependent perturbation responses. ChromaPert combines molecular and DNA-informed locus priors with paired control RNA–ATAC features to jointly represent target identity and measured cellular context. Its Chromatin-Guided Response Router (CGRR) builds a source bank of control-derived expression offsets and retrieves relevant sources using perturbation/locus or control-ATAC similarity. These offsets account for baseline differences, while conditional transport flow learns the remaining response. Across three unseen-target and held-out-state settings on K562 CAT-ATAC and Perturb-Multiome, ChromaPert achieves the highest all-gene response correlation among evaluated methods. When transferring known perturbations to held-out states, it improves correlation by 23.2% across all genes and 46.2% for the twenty most responsive genes over the strongest respective baselines. Correct RNA–ATAC pairing improves recovery of chromatin-associated response slopes over shuffled pairing, while cross-state predictions retain state-specific responses to the same perturbation.

AlphaFold 3 predicts protein structures with remarkable accuracy, yet how structural information emerges within the model remains poorly understood. Here, through causal interventions on internal representations and direct probing of every Pairformer block, we trace the formation of global protein geometry and identify the multiple sequence alignment (MSA) as a structural shortcut to the fold. Removing the MSA largely preserves local secondary structure while disrupting the long-range relationships that define global topology. Restoring the MSA-enriched pair representation at only forty residues recovers most of this lost organization, including at pairs never directly modified. This contribution depends on the detailed direction of the MSA module's output rather than its magnitude. The Pairformer rapidly converts this signal into global geometry: the final fold becomes recoverable by approximately block 9 of 48 for a majority of proteins, roughly twenty-seven blocks before the model's decoder can render it, whereas without the MSA it remains inaccessible for most proteins throughout the pass. Which homologs are supplied shapes this trajectory more strongly than which query is supplied; it persists for a designed query that never evolved but collapses for a shuffled sequence. Most importantly, an alignment built for a different protein that shares the fold, supplied only at the structurally corresponding columns, raises the median TM-score against experiment from 0.44 to 0.72, while the same alignment shifted a few residues along the chain performs worse than supplying no alignment at all. What AlphaFold 3 reads from an alignment is therefore a description of the fold itself, transferable between proteins that share one, rather than the query's own evolutionary history. This explains both its accuracy and the limits of what it has solved.

Petri nets have been used to describe chemical processes such as reactions.They map well to chemistry: Places are the bonds between atoms and the free valence of each atom, a token is a unit of bond order, a transition forms or breaks a bond, the conserved quantities are the valence budgets of the atoms, and the enabling rule is the valence rule. These semantics are not guaranteed by learned models of reactions or neural networks that are built on Petri nets that use the net as a scaffold for message passing. Here, we ask what architecture remains a Petri net for every value of its weights. We find the answer in the theory, where all semantics of a net share the firing form $m^\prime=m+C\sigma$, locality, as enabling reads only the inputs of a transition, and the enabling rule, and we prove that conservation forces the firing form and that non-negativity forces the enabling rule on local rate laws. This leaves free the rate law, which is the propensity of each transition to fire. We introduce Neural Petri Flow, which learns this rate law, or a readout for classification, and hard-wires the rest as parameter-free layers. On what we denote a valence net, atom mapping, reaction classification, and forward prediction become three tasks on one firing vector. Without training, the minimum firing vector maps 88.8% of the curated Golden set against 85.6% for RXNMapper, and 88.7 against 77.9% of the enzymatic reactions of EnzymeMap. On USPTO-480K, NPF trained on these firing vectors predicts 87.7% of the products and 67.4% when trained on a 1% subset of the training reactions. EC numbers of ECREACT are predicted at the third level for 90.2% of reactions, 5.6 points ahead of the best published method. With electrons as tokens, the same token game predicts 90.5% of the elementary steps of FlowER first, ahead of the published baseline, and every top-1 prediction is a valid molecule without a filter.