Quantitative Biology

Machine learning models trained on biochemical data are routinely evaluated using splits that fail to account for relational structure, causing information leakage and over-optimistic performance estimates. Existing splitting methods lack theoretical grounding and scale at best quadratically. We introduce the Relational Generative Process (RGP), a mathematical formalization explaining why relational structure arises in biochemical datasets, and Refnd, a splitting algorithm that leverages a proximity graph computed in loglinear time using Hierarchical Navigable Small World (HNSW). We validate on an antimicrobial peptide dataset, showing that Refnd splits yield lower but more realistic evaluation performance than traditional splits. Refnd is applicable to any dataset arising from an RGP such as protein sequences and structures, small molecules, and nucleotide sequences, and is openly available as a Rust accelerated Python package: pip install refnd.

High-content morphological profiling (Cell Painting) yields sensitive, high-dimensional signatures of cellular state, but translating longitudinal morphology trajectories into interpretable biology remains difficult, especially for weak, chronic perturbations such as low-dose-rate ionizing radiation. Large language models (LLMs) can synthesize heterogeneous evidence into biological narratives, yet their scientific use requires quantitative auditing. We present an evaluation-first, retrieval-augmented interpretation framework for longitudinal Cell Painting morphology, applied to a 9-week RPE-1 time course across five dose rates (0.003--6.0 mGy/hr). Week-matched treated-control morphology deltas are combined with retrieved perturbation neighbors, pathway context, and literature evidence through stable evidence identifiers, enabling an LLM to generate structured, evidence-linked hypotheses that are hierarchically summarized while preserving provenance. We introduce two quantitative auditing tests: V1 citation validity, which verifies that cited evidence identifiers exist in the prompt, and V2 proxy-based morphology compatibility, which evaluates consistency between predicted biological processes and the most altered morphology features. In our experiments, V1 detected no invalid evidence references, while V2 showed meaningful morphology compatibility that increased with perturbation strength and was positively associated with an independent morphology drift summary. The framework produces auditable, falsifiable biological hypotheses, including an adaptive phenotype involving metabolic reprogramming and proteostatic stress at lower dose rates (0.003--0.3 mGy/hr). Current limitations include proxy-based evaluation and the lack of ground-truth mechanism labels.

Brain-computer interface neurofeedback (BCI-NFT) has shown promise for neuromotor rehabilitation, but its clinical adoption -- particularly in pediatric populations -- remains limited in part by the lengthy calibration required before each therapy session. This study developed and evaluated a deep-learning framework to reduce calibration requirements for movement-related cortical potential (MRCP)-based movement-intention detection in children with cerebral palsy (CP). Electroencephalography (EEG) was collected during repeated ankle dorsiflexion tasks across 27 sessions in four children with CP. A bidirectional long short-term memory (Bi-LSTM) network was evaluated using seven training strategies, ranging from conventional within-session calibration to cumulative cross-subject learning and transfer learning. Cross-subject cumulative learning achieved 91\% accuracy without within-session calibration, while the addition of transfer learning increased accuracy to 93\% with minimal within-session calibration. Both approaches significantly outperformed conventional calibration strategies and achieved the highest F1-scores and receiver operating characteristic (ROC) performance, demonstrating robust generalization across sessions and participants. These findings show that cumulative learning and transfer learning can substantially reduce calibration requirements while maintaining high decoding performance, supporting the development of clinically practical MRCP-based pediatric BCI systems

Single-cell datasets are increasingly costly to store, audit, and reuse for model training. Dimensionality reduction and dataset distillation can reduce this burden, but conventional distillation methods often produce synthetic expression profiles that cannot be traced to an assayed cell. We formulate traceable single-cell data distillation as retaining original cell identifiers and gene symbols under fixed cell and gene budgets. The resulting training subset remains connected to measured counts, labels, and assay metadata, so unexpected predictions can be checked against their source data. We propose two real-cell selectors. Fixed-CF uses static characteristic-function matching. Minmax-CF solves an entropy-regularized discrete min--max problem that upweights poorly preserved directions and adds only observed cells. Across donor-, technology-, and perturbation-level shifts on three datasets, Minmax-CF retains 96.52% of Full balanced accuracy on MS, approximately matches Full on average on hPancreas with a median $2.55\times$ GPU speedup in the all-gene setting, and obtains the lowest pathway error among compressed methods on Norman. Performance remains weaker for rare states, some technology shifts, unseen perturbation components, and settings where fidelity is weakly associated with downstream utility. Because the selected IDs refer to measured cells, these cases can be investigated by inspecting the corresponding training support, labels, and assay metadata. Minmax-CF consistently reduces worst-direction discrepancy, while downstream utility and cost vary across datasets and tasks.

Genomic language models achieve strong performance across regulatory-genomics tasks, yet what these models internally represent remains opaque, and the field lacks a principled procedure for verifying that an apparent ``concept'' inside a model is real rather than an artifact of sequence composition. We introduce a framework that combines sparse dictionary learning with causal intervention to extract, validate, and causally test interpretable features in genomic foundation models. Training top-$k$ sparse autoencoders on the hidden activations of two architecturally distinct models, Nucleotide Transformer ($6$-mer tokenization) and DNABERT-2 (byte-pair encoding), we recover thousands of monosemantic features that map to transcription-factor (TF) sequence motifs. We show that the naive validation of such features against position weight matrices is severely confounded by GC composition and repetitive elements, producing hundreds of spurious ``TF features'', and we develop a composition-matched, binding-resolved protocol that removes these confounds. Critically, we move beyond correlation: by ablating individual dictionary directions during the model's forward pass and measuring the induced shift in the model's own predictive distribution, we establish that specific features are \emph{causally} used to represent cell-type-specific TF binding, not merely motif presence. Across three transcription factors (CTCF, GATA1, REST) and both architectures, causally validated binding features emerge reproducibly ($7$--$14$ of $15$ tested features per condition), while two classes of negative control, scrambled binding labels and randomly selected features, yield no detectable signal. The framework is purely computational, uses only public data, and provides a reusable standard for interpretability claims in genomic deep learning.

The visually perceived vertical is biased by the orientation content of the visual field. Li and Matin (2005a, 2005b) reported three regularities of this induced vertical (VPV): a roll-tilted peripheral line shifts it about linearly with orientation; two lines combine about linearly; and symmetric tilts cancel. We show all three follow from one principle. The induced vertical is half the argument of the first circular moment of stimulus orientation in the doubled-angle domain, $\phi=\frac{1}{2}\arg c_1$ with $c_1=\sum_j A_j e^{i2\gamma_j}$ (each line counted at twice its angle). Equivalently it is the principal axis of the orientation structure tensor, or an orientation population vector; we call it the orientation order-parameter model (PLUMB). The angle-doubling is forced: orientation is a director ($\theta\equiv\theta+\pi$), so a circular-mean readout must be the doubled-angle one, and linear tracking, combination, and cancellation are signatures of any such readout, not separate findings. Li and Matin's 2-, 3-, and 4-line combination coefficients are then closely matched with no per-configuration free parameters in the angles; the magnitudes come from one mass-action length function (their Fig. 5, three constants). The data refute complete summation but do not separate the model from simple averaging for long lines; the decisive parameter-free tests are the short-line regime and a $|\cos 2\theta|$ strength law with a null at $45^\circ$. Read as a sum and a difference across the two hemifields, the same order parameter yields perceived vertical and eye level, recovering the reversed rules of Shavit, Li and Matin (2013); their 30-observer trial data confirm the predicted sub-additive cross-field combination. The account is stimulus-side and image-computable, linking induced vertical and eye level to classical image-orientation descriptors.

Microbial communities are ubiquitous in nature but how they grow on available resources is still poorly understood. Communities are complex systems harboring thousands of microbial species that interact through resource competition. The classical MacArthur consumer-resource model has been shown to underestimate formed community biomass. A recent new microbial consumer-resource model (MiCRM) considers the inclusion of inter-specific interactions mediated by metabolite exchange (cross-feeding) where the various bacterial growth byproducts can be reused by other species for their own growth. We study persistence, feasibility and stability for MiCRM under some simplifying assumptions using slow-fast approximation. We show e.g. the non-persistence of the microbial community when the number of resource species M is smaller than the number of consumer species S. We then study the stability of the slow steady state when the number or survivors S is smaller than M, and show that such equilibria are generically stable. We finally propose a stochastic slow-fast version of the model having fast Poisson steady state and study related extinction events.

The logistic equation treats the definition of its variable freely. It is undoubtedly the population density. The logistic equation in the differential version requires continuity of processes occurring in the population. In the differential version, we do not know what happens to the population between time steps. When filling these gaps, models with features different from the logistic equation appear. If we have a micro-scale model describing the fate of individuals and interactions between them, then the macro description for this model is not the equation for population density (e.g. the logistic equation) but the model of matter circulation or energy flow through this system. We lack mathematics to describe purely biological processes.

Mechanistic ordinary differential equation (ODE) models are widely used in systems biology, but uncertainty quantification (UQ) remains difficult when only a subset of state variables is experimentally observed. Existing Bayesian and likelihood-based approaches can be computationally demanding for nonlinear, weakly identifiable, or high-dimensional systems. We present a framework, and its corresponding software CUQDyn1 Plus, for UQ in partially observed ODE systems. Our method combines leave-one-out jackknife+-style empirical calibration for observed states with sensitivity-based Gaussian uncertainty propagation for hidden states. The software supports global parameter estimation, covariance propagation, bootstrap trajectory uncertainty and simulation-based calibration. It also facilitates comparison with Bayesian workflows, automated reporting, and reproducibility diagnostics. Validation on six benchmark systems shows accurate behavior in well-conditioned cases and model-dependent degradation under nonlinearity, weak identifiability, or global branch-switching non-identifiability. CUQDyn1 Plus provides a practical and computationally efficient UQ workflow for systems biology models with observed and latent states. Its diagnostic outputs help identify when local Gaussian propagation is reliable and when uncertainty bands should be interpreted cautiously, making it a useful complement to fully Bayesian workflows.

Antibodies are essential proteins that play a central role in immune recognition by binding specific antigen molecules. Although recent protein language models have enabled progress in single-chain protein modeling and generation, they often fall short in antigen-specific antibody design, where effective modeling requires explicit pairing between antibody and antigen, particularly at the epitope level. To address these limitations, we introduce AAMFM, an Antigen-specific Antibody Multimodal Foundation Model that learns unified representations of antibody sequences and structures conditioned on antigen context. AAMFM incorporates rich antigen information including geometric interfaces and epitope annotations via a cross-modal adapter, enabling joint modeling of antibody-antigen interactions in a shared latent space. To further guide the model toward functional relevance, we fine-tune AAMFM using Calibrated Direct Preference Optimization (Cal-DPO), leveraging preference signals extracted from a strong structural prior to align learning with binding-specific objectives. Extensive experiments demonstrate that AAMFM achieves state-of-the-art performance in functional antibody design, revealing its potential for antigen-specific antibody engineering. Our code is available at this https URL.

The rapid growth of biomedical knowledge has made the validation of automatically generated biological annotations a major bottleneck in biomedical curation. While computational methods can rapidly produce large numbers of candidate annotations, determining which are biologically valid still requires costly expert review. Prioritizing these candidates before manual curation has therefore become a fundamental challenge. Machine learning techniques can support this process by exploiting biomedical knowledge graphs (bioKGs), which capture biological entities and their functional associations. In this work, we propose a framework that leverages bioKGs to estimate the plausibility of candidate annotations and guide expert curation. Starting from knowledge graph embeddings, we train relation-specific binary classifiers using a community-based negative sampling strategy to obtain reliable confidence estimates. We then introduce a family of plausibility measures that combine classifier confidence, classifier reliability, and the semantic context provided by alternative relationships involving the same pair of biological entities. Unlike conventional confidence estimation, the proposed approach explicitly accounts for multiple biologically meaningful relations that may coexist between the same entities. Experimental results on five large bioKGs demonstrate that the proposed negative sampling strategy consistently improves classifier robustness, increasing balanced accuracy by an average of 5.8%. Moreover, the plausibility measures outperform classifier confidence alone, enabling more effective prioritization of candidate annotations for expert review. Overall, our results show that the use of bioKGs improves the efficiency of AI-assisted biomedical curation while preserving expert control over the final annotation assessment.

High-dimensional biological datasets with molecular, spatial, and temporal dimensions are increasingly common. However, their analysis requires approaches that can integrate multiple data axes, accommodate noisy and missing measurements, and capture both dynamic interactions and localization changes. To address this, we provide an end-to-end tensor analysis pipeline that handles sparsity by employing tensor decomposition methods (HOSVD and CP) that are augmented with a binary mask for missing data and a framework for measurement error. We showcase this on a three-dimensional mammalian lipid transport dataset depending on lipid identities, organelle localizations, and time-series abundances. Our approach successfully identifies specific lipid-organelle pairs undergoing rapid temporal evolution, uncovers modules of lipids that co-vary across organelles and time, and extracts latent factors representing global redistribution trajectories. Direct comparison with a previous kinetic ODE model confirms that the tensor decompositions faithfully reproduce key lipid flux features.

Recent work established that under active inference, linear-Gaussian state-space models lose their epistemic drive (any incentive to act so as to gain information) "under any circumstances". The epistemic term of the Expected Free Energy becomes constant: the agent flattens to a Kalman filter whose gain sequence is fixed in advance, regardless of action. The minimal departure that restores the drive is unknown; the only established route is control entering the dynamics multiplicatively; the observation side of this boundary is unexplored. We show that state-dependent observation noise is such a departure: a covariance R(x) that varies with the state x, representing a sensor's accuracy degrading with range. The agent runs the standard first-order Gaussian filter of this literature, R evaluated at the predicted mean. Coupling R(x) to a controllable latent mean makes the posterior covariance, and hence the effective Kalman gain, depend on the action. Consequently, no fixed linear-Gaussian filter reproduces the agent and, under a mild rank condition on the observation map and a non-degeneracy condition on R(x), epistemic value is no longer constant; for scalar observations, reachable non-constancy alone is needed. This is a minimal constructive instance of the Bar-Shalom-Tse dual effect in the agent's maintained covariance: actions now influence the quality of future estimates, not merely the state. Our library cpomdp detects the incompatibility from model specification alone and raises a typed IncompatibleLinearizationError. The theorem ships with an executable witness: exhibiting any fixed filter that reproduced the agent's beliefs would refute both theorem and witness at once. Together this offers a precise, observation-side characterisation of curiosity in a Gaussian agent, bridging dual control and active inference.

Recent works have highlighted the use of dual nostril 'stereo' olfaction by a variety of animals. In this work we perform "back of the envelope" calculations to determine when stereo olfaction is useful compared to single nostril 'mono' olfaction. We find that stereo olfaction is advantageous when there are large relative changes in the odour concentration, and when the spatial length scales of correlations in air are large, such as in the boundary layer near surfaces. In other words, stereo olfaction is useful when animals are searching surfaces for olfactory edges, such as when tracking odour trails.

Subjective experience is central to psychological science, yet methods for studying it force a choice between depth and scale. Classical Experience sampling, as in ecological momentary assessments (EMA), captures experience as it occurs, but it confines participants to predetermined response formats that prescribe how experience is measured. Descriptive Experience Sampling (DES) instead investigates specific moments in depth through expert expositional interviews, but its reliance on scarce trained interviewers keeps samples small. Large language model (LLM) systems can scale qualitative interviewing. Some models operationalize established interviewing methods such as motivational interviewing, yet none is grounded in a method for apprehending inner experience. Here we present an AI interviewer that aspires to operationalize DES into an explicit, inspectable reasoning architecture. At each turn it appraises the participant's message across eleven quality dimensions, maintains a conservative account of what has been established, selects a stage-appropriate intervention, and composes a single non-leading query, always holding that temporal grounding precedes experiential content. It was derived from the full corpus of DES transcripts and refined with the method's originator Russell T. Hurlburt. To our knowledge it is the first AI interviewer grounded in an established method for studying inner experience. The interviewer runs inside Introscope, an application that delivers the beeps and conducts the interviews and a study platform that lets researchers run studies via shareable links and review the sampled experience. It is demonstrated in an accompanying video this https URL. Pending validation studies, we will make it freely available to researchers and the public, for crowdsourced sampling and individual exploration of inner experience.

AI-assisted colonoscopy systems commonly report frame-level confidence scores, but confidence alone does not show whether a prediction is spatially plausible, temporally persistent or reliable under degraded image quality. We propose EndoExplain, a lightweight and reproducible audit framework for endoscopic AI review. The framework integrates classification confidence, lesion segmentation, CAM-style visual attribution, attribution-mask alignment, frame-quality indicators and temporal event summarisation, with the aim of separating signals that are often conflated in computer-aided detection pipelines. On HyperKvasir, the selected EfficientNet-B0 classifier reaches 0.9280 test accuracy over ten endoscopic classes and 0.9969 ROC-AUC for the polyp_family versus rest view. The selected U-Net++ EfficientNet-B1 segmenter reaches Dice 0.9318 and IoU 0.8826 on the segmented test split. A strict top-20% multi-method attribution audit shows that attribution method strongly changes explanation-mask agreement: Eigen-CAM gives the strongest overlap, whereas Grad-CAM++ remains weakly aligned. A frozen-model external sanity check on ETIS-LaribPolypDB and CVC-ClinicDB preserves this attribution-method ranking while showing dataset-dependent overlap. A human-reviewed clip-level temporal benchmark over 60 HyperKvasir videos reaches event F1 0.8081 at the pre-specified threshold 0.85 using one-to-one event matching with overlap >= 1 s. A clinician-informed external plausibility review supported the clinical readability of separating confidence, localisation, attribution, quality metadata and temporal context. The resulting cockpit-style review layer presents these signals as distinct auditable outputs. This is a retrospective research prototype and not a clinically validated medical device.

Pre-trained foundation models (FMs) have begun transforming single-cell genomics, but scaling them raises privacy concerns. Moreover, unlike text data, single-cell data is unordered and exhibits a unique tabular structure that current single-cell FMs overlook. We introduce Tabula, a privacy-preserving FM designed with federated learning (FL) that explicitly models the tabular structure of single-cell data. To deploy Tabula, we further developed Chiron, a decentralized AI agent-enabled platform for collaborative training across institutions without sharing raw data. Beyond strong performance across downstream benchmarks, Tabula reveals combinatorial regulatory logic across diverse biological systems, including hematopoiesis, pancreatic endogenesis, neurogenesis, and cardiogenesis. Using a new scRNA-seq dataset of paired young and aged human fibroblasts, Tabula nominates rejuvenation factors through age- and identity score-guided in silico prioritization, outperforming conventional approaches. Thus, Tabula represents an important advance in single-cell foundation modeling by integrating tabular learning with FL, paving the way toward privacy-preserving virtual cells for human health.

We discuss the problem of limit cycles in an aggregate-type models of liquidity growth proposed and studied both theoretically and numerically by W. Semmler and M. Sieveking in \cite{SemmlerSieveking}. Their model is locally of predator-prey type with cannibalism in predator (profit) and logistic bound on prey's population (liquidity). We modify the model to include the weak Allee effect in liquidity and to study the local bifurcations of equilibria and limit cycles by means of the Hopf-Andronov theory. The main motivation for this study follows the economic theory of the so-called corridors of stability introduced by Leijonhuvud in \cite{Leijonh}. Similar importance of these effects is, however, found also in the originally developed dynamical models in biology, especially in mathematical ecology.

Markov state models (MSMs) have become ubiquitous tools for analyzing molecular dynamics (MD) simulations because of their simple, powerful premise: although complete MD sampling may be impossible, the MSM can "stitch together" transition probabilities derived from local sampling to provide a global picture of kinetics and mechanisms. In the standard MSM framework, the available MD data is organized into a single transition matrix, which is then used to estimate all observables at a lag time chosen so the coarse-grained dynamics are approximately Markovian. This approach leads to avoidable model bias and motivates long lag times that obscure short-timescale processes of interest. In contrast, this paper shows how to obtain unbiased coarse-grained observables at any fixed lag time and for any fixed coarse-graining in the limit of infinite, properly weighted data. The central idea is to replace the single-matrix framework with two transition matrices -- one representing equilibrium dynamics and another representing source-sink recycling dynamics -- and use the correct matrix or matrices to estimate the matched dynamical observables.

The shape of a planar curve is the geometric information that remains once translation, rotation, scale and reparametrisation are removed and is of interest in many health applications, e.g. in neuroimaging. We propose a deep shape regression model for open planar curves that admits multimodal and high-dimensional covariates. Representing curves as complex-valued functions, we show that the conditional full Procrustes mean is the leading eigenfunction of the conditional covariance. To estimate this covariance surface, we propose a novel deep conditional covariance smoother with modality-specific encoders - e.g. splines for scalar covariates and convolutional networks for images, which classical spline smoothers cannot accommodate. Our model is by construction invariant to the translation, rotation and scaling of the input curves and handles sparsely and irregularly sampled curves. We further provide an algorithm for elastic mean estimation that also removes parametrisation by iterating covariance smoothing, rotational alignment and parametrisation alignment. We illustrate the method on simulated outlines with known conditional mean and multimodal covariates, and give a first application to hippocampal outlines from the ADNI cohort, recovering covariate effects consistent with the literature. Code is available at this https URL.

AI researchers describe state-of-the-art models as one thing repeated at scale: the Transformer, wired identically for text, pixels, or speech. Neuroscientists describe the cortex as a mosaic - dense Layer 4 in visual cortex for spatial encoding, thick Layers 5/6 in motion cortex for temporal integration - different jobs solved by different structures. This paper argues the gap is a structural error, not a stylistic one, and is measurable. A century of cytoarchitecture, from Brodmann to single-cell Patch-seq, shows distinct cognitive functions are implemented by qualitatively different structures, not by rescaling one template. The convolutional neural network is the field's own proof: local receptive fields and hierarchical depth encoded this prior directly, reaching strong image recognition on far less data than later architectures needed. The paper traces how this lesson was discarded: the "Hardware Lottery" made the Transformer the path of least resistance, not the principled choice, and Mixture-of-Experts, often cited as diversity, in fact partitions parameters among identical experts. A functionalist analysis shows the Transformer is best understood as a functional analog of the hippocampal formation, not a general-purpose cortex - the same mistake as treating cortex as one giant Broca's area, except the field has now standardized on a giant hippocampus, applied to tasks it was never built for: audition, executive gating, working memory. The paper closes with an alternative: a Heterogeneous Topological Network, a System of Systems in which distinct modules keep the inductive bias their computation demands and communicate through standardized interfaces. This is a design discipline for AI architects, not cognitive science: specify modularity before training, using structural evidence as a design input rather than reverse-engineering architecture from a trained model's behavior.

The automated reconstruction of patient family history is a common challenge in genetic counseling for disease prevention. Such a family history is usually determined for a particular subset of diseases that are Mendelian, i.e. monogenic, and classified into three categories depending on the chromosome the gene is located: autosomal, X-linked or Y-linked. Mendel's inheritance laws allow for simple probabilistic modeling of the genetic transmission of monogenic disorders. Genetic counsellors use knowledge about the patient's family history and Mendelian laws for assessing risks of transmitting or inheriting congenital conditions. We present this http URL, a probabilistic logic programming algorithm in ProbLog for deriving probabilities of inheritance of genotypes and phenotypes for genes with two alleles through multiple generations. In particular, the user can input genotypes and phenotypes for a patient and its family, and automatically determine the most probable genetic family history. We illustrate the ProbLog model on practical examples of patient pedigrees from the literature and from a genetic counseling handbook. We show that our method correctly infers probability of individual genotypes from knowledge about familial genotypes, yielding the same results as tool pedprobr. However, unlike pedprobr, our approach can exploit knowledge about familial phenotypes. It can also directly distinguish between autosomal, X-linked, and Y-linked disorders, using its intuitive logical modelling. We provide our ProbLog tool for free and open-source on GitHub, making it easily available for genetic counsellors. We conclude on the importance of providing explainable formal methods for a task that clinicians might want to perform using proprietary software.

Nanopore devices have emerged as powerful tools for single-molecule sensing, with potential for rapid, portable diagnostics. They detect changes in ionic current as analytes enter nanometer-scale pores, providing a means of identifying diverse biomarkers from their characteristic signal patterns. However, these signals are highly complex, and reliably assigning them to specific molecules remains a major challenge. Here, we address this by introducing a multi-modal deep learning architecture that jointly processes multiple signal representations, including raw time-series data, wavelet-based images, and static feature vectors. Our approach surpasses existing methods by more than 10 percentage points on a 42-peptide benchmark and transfers to a 20-amino-acid dataset with near-perfect accuracy. The model integrates complementary information from these representations, with attention analysis showing that the time-series and wavelet-image inputs emphasize different features of the same event. Together, these results demonstrate the potential of machine learning to enable robust, high-accuracy molecular identification with nanopore sensors.

Clinical biomarker workflows in translational research settings often rely on spreadsheet-driven tracking, manual quality control (QC) reconciliation, and loosely integrated systems, resulting in limited state visibility, delayed reporting, and increased operational risk. These challenges are particularly pronounced in multi-day assays such as Luminex-based quantification of Fragile X Messenger Ribonucleoprotein (FMRP), where HIPAA-compliant data governance, deterministic workflow progression, and coordinated communication across laboratory and clinical teams are required. This paper presents FMRP-LEAN, a HIPAA-compliant, AI-augmented Laboratory Information Management System (LIMS) architecture that formalizes biospecimen lifecycle management through a finite-state workflow model with explicit transition guards and dwell-time observability. The system integrates a self-hosted Supabase/PostgreSQL stack deployed within hospital-controlled infrastructure, hybrid edge-internal isolation with encrypted tunneling and loopback-only services, and bi-directional REDCap synchronization. A unified MRN-UUIDv7 identifier framework with QR-based tracking ensures traceable clinical-research linkage under PHI residency constraints. FMRP-LEAN incorporates automated statistical QC pre-screening and a governance-constrained AI operations module that operates exclusively on aggregate projections, with deterministic fallback guarantees. Deployment demonstrates improved workflow observability, reduced QC latency, and enhanced cross-role transparency between laboratory technicians, research coordinators, and patient-facing teams. The architecture provides a reproducible model for secure, state-explicit, and AI-augmented clinical research workflows in regulated healthcare environments.

In network neuroscience, functional brain systems are often characterized using separate yet related graph-theoretic or spectral descriptors, overlooking how these properties covary and partially overlap across individuals and conditions. We anticipate that dense, weighted functional connectivity graphs lie on a low-dimensional latent geometry along which both topological and spectral structures vary smoothly at the population level. Although graph-based deep learning offers a powerful framework for modeling these brain connectomes, supervised approaches are constrained by the limited availability of labeled data. Existing unsupervised graph representation methods also typically focus on node-level embeddings, which are limited in capturing compact graph-level representations that preserve information from dense functional connectomes. To address these gaps, we learn compact brain graph representations using a graph transformer autoencoder, where domain-specific, aligned functional gradient geometry provides an inductive bias to guide learning. Despite being trained in a fully unsupervised manner, our approach meaningfully separates cognitive states and enables decoding of visual stimuli, with performance further improved by incorporating neural dynamics. In parallel, to enable generation of synthetic brain graphs, we fit a diffusion model to the learned latent representation and decode samples back to dense connectomes.

Understanding how natural selection unfolds across space and time is a central problem in evolutionary biology. Classical models such as the Moran process capture stochastic birth-death dynamics in structured populations, while reaction-diffusion equations like the Fisher-Kolmogorov-Petrovsky-Piskunov (FKPP) equation describe deterministic wavelike spread. In this work, we bridge these perspectives by deriving partial differential equations for the spatiotemporal limit of Moran dynamics in continuous media. Our model incorporates two distinct fitness components: fecundity (birth rate) and viability (death rate). We demonstrate that the resulting selective wave speeds differ substantially in spatial Moran birth-death (BD), Moran death-birth (DB), and FKPP dynamics. When fecundity drives the dynamics, we observe that the selective waves decelerate for the BD process, whereas in the DB process the wave propagates with a higher, constant speed. In contrast, when viability drives the process, the DB wave accelerates, while the BD and FKPP waves maintain comparable constant speeds. We extend the framework to heterogeneous media, represented as weighted lattice graphs in one or two dimensions. We derive a continuous-space analog of isothermal graphs and establish that the isothermality condition corresponds to the conservation of a local current.

Sometimes local search algorithms cannot efficiently find even local peaks. To understand why, I look at the structure of ascents in fitness landscapes from valued constraint satisfaction problems (VCSPs) parameterized by the treedepth of their constraint graphs. There are existing constructions of VCSPs with logarithm treedepth that represent fitness landscapes where all ascents are exponential from some initial assignment. I improve these bounds by showing that with loglog treedepth, superpolynomial ascents exist; and for polylog treedepth, there are initial assignments from which all ascents are superpolynomial. My hope is that these examples of sparse VCSPs can help us better understand the barriers to efficient local search.

State-space models are dynamical systems defined by a latent and an observed process. In ecology, stochastic state-space models in discrete time are most often used to describe the imperfectly observed dynamics of population sizes or animal movement. However, several studies have observed identifiability issues when state-space models are fitted to simulated or real data, and it is not currently clear whether those are due to data limitations or more fundamental model non-identifiability. To investigate such theoretical identifiability, a suitable exhaustive summary is required, defined as a vector of parameter combinations which fully determines the model. Previous work on exhaustive summaries has used expectations of the stochastic process, so that noise parameters are unaccounted for. In this paper, we build an exhaustive summary using the spectral density of the observed process, which fully accounts for all mean and variance parameters. This diagnostic is applied to contrasted ecological models and we show that they are generally theoretically identifiable, unless some model compartements are unobserved. This suggest that issues encountered while fitting models are mostly due to practical identifiability.

Pharmaceutical drug discovery demands machine learning (ML) infrastructure that goes beyond general-purpose Machine Learning Operations (MLOps): inference-time composition of multiple models for multi-parameter optimization (MPO), version management for physics-based models without serialized ML artifacts, enterprise compound library precomputation, and governance structured around scientific organizational units rather than generic access controls. No existing commercial or open-source platform simultaneously addresses this full set of requirements. This paper presents the Model Gateway, a cloud-based platform for managing machine learning and scientific computational models across drug discovery pipelines, providing centralized version control, pharma-structured governance, asynchronous execution, consensus model orchestration, automated retraining, and a unified application programming interface (API) service for heterogeneous clients including molecular design suites and Large Language Model (LLM) agents. In production at Eli Lilly, the platform governs more than 200 deployed models spanning small molecule, peptide, and antibody modalities and serves more than five downstream applications across all phases of the Design-Make-Test-Analyze cycle.