Interpretable (not just posthoc-explainable) medical claims modeling for discharge placement to reduce preventable all-cause readmissions or death.

We developed an inherently interpretable multilevel Bayesian framework for representing variation in regression coefficients that mimics the piecewise linearity of ReLU-activated deep neural networks. We used the framework to formulate a survival model for using medical claims to predict hospital readmission and death that focuses on discharge placement, adjusting for confounding in estimating causal local average treatment effects. We trained the model on a 5% sample of Medicare beneficiaries from 2008 and 2011, based on their 2009-2011 inpatient episodes (approximately 1.2 million), and then tested the model on 2012 episodes (approximately 400 thousand). The model scored an out-of-sample AUROC of approximately 0.75 on predicting all-cause readmissions-defined using official Centers for Medicare and Medicaid Services (CMS) methodology-or death within 30-days of discharge, being competitive against XGBoost and a Bayesian deep neural network, demonstrating that one need-not sacrifice interpretability for accuracy. Crucially, as a regression model, it provides what blackboxes cannot-its exact gold-standard global interpretation, explicitly defining how the model performs its internal "reasoning" for mapping the input data features to predictions. In doing so, we identify relative risk factors and quantify the effect of discharge placement. We also show that the posthoc explainer SHAP provides explanations that are inconsistent with the ground truth model reasoning that our model readily admits.

Gradient-flow adaptive importance sampling for Bayesian leave one out cross-validation for sigmoidal classification models.

We introduce a set of gradient-flow-guided adaptive importance sampling (IS) transformations to stabilize Monte-Carlo approximations of point-wise leave one out cross-validated (LOO) predictions for Bayesian classification models. One can leverage this methodology for assessing model generalizability by for instance computing a LOO analogue to the AIC or computing LOO ROC/PRC curves and derived metrics like the AUROC and AUPRC. By the calculus of variations and gradient flow, we derive two simple nonlinear single-step transformations that utilize gradient information to shift a model's pre-trained full-data posterior closer to the target LOO posterior predictive distributions. In doing so, the transformations stabilize importance weights. Because the transformations involve the gradient of the likelihood function, the resulting Monte Carlo integral depends on Jacobian determinants with respect to the model Hessian. We derive closed-form exact formulae for these Jacobian determinants in the cases of logistic regression and shallow ReLU-activated artificial neural networks, and provide a simple approximation that sidesteps the need to compute full Hessian matrices and their spectra. We test the methodology on an n≪p dataset that is known to produce unstable LOO IS weights.

A mathematical model of SARS-CoV-2 immunity predicts paxlovid rebound.

Nirmatrelvir/ritonavir (Paxlovid), an oral antiviral medication targeting SARS-CoV-2, remains an important treatment for COVID-19. Initial studies of nirmatrelvir/ritonavir were performed in SARS-CoV-2 unvaccinated patients without prior confirmed SARS-CoV-2 infection; however, most individuals have now either been vaccinated and/or have experienced SARS-CoV-2 infection. After nirmatrelvir/ritonavir became widely available, reports surfaced of "Paxlovid rebound," a phenomenon in which symptoms (and SARS-CoV-2 test positivity) would initially resolve, but after finishing treatment, symptoms and test positivity would return. We used a previously described parsimonious mathematical model of immunity to SARS-CoV-2 infection to model the effect of nirmatrelvir/ritonavir treatment in unvaccinated and vaccinated patients. Model simulations show that viral rebound after treatment occurs only in vaccinated patients, while unvaccinated (SARS-COV-2 naïve) patients treated with nirmatrelvir/ritonavir do not experience any rebound in viral load. This work suggests that an approach combining parsimonious models of the immune system could be used to gain important insights in the context of emerging pathogens.

Distributing task-related neural activity across a cortical network through task-independent connections.

Task-related neural activity is widespread across populations of neurons during goal-directed behaviors. However, little is known about the synaptic reorganization and circuit mechanisms that lead to broad activity changes. Here we trained a subset of neurons in a spiking network with strong synaptic interactions to reproduce the activity of neurons in the motor cortex during a decision-making task. Task-related activity, resembling the neural data, emerged across the network, even in the untrained neurons. Analysis of trained networks showed that strong untrained synapses, which were independent of the task and determined the dynamical state of the network, mediated the spread of task-related activity. Optogenetic perturbations suggest that the motor cortex is strongly-coupled, supporting the applicability of the mechanism to cortical networks. Our results reveal a cortical mechanism that facilitates distributed representations of task-variables by spreading the activity from a subset of plastic neurons to the entire network through task-independent strong synapses.

Multiple models for outbreak decision support in the face of uncertainty.

Policymakers must make management decisions despite incomplete knowledge and conflicting model projections. Little guidance exists for the rapid, representative, and unbiased collection of policy-relevant scientific input from independent modeling teams. Integrating approaches from decision analysis, expert judgment, and model aggregation, we convened multiple modeling teams to evaluate COVID-19 reopening strategies for a mid-sized United States county early in the pandemic. Projections from seventeen distinct models were inconsistent in magnitude but highly consistent in ranking interventions. The 6-mo-ahead aggregate projections were well in line with observed outbreaks in mid-sized US counties. The aggregate results showed that up to half the population could be infected with full workplace reopening, while workplace restrictions reduced median cumulative infections by 82%. Rankings of interventions were consistent across public health objectives, but there was a strong trade-off between public health outcomes and duration of workplace closures, and no win-win intermediate reopening strategies were identified. Between-model variation was high; the aggregate results thus provide valuable risk quantification for decision making. This approach can be applied to the evaluation of management interventions in any setting where models are used to inform decision making. This case study demonstrated the utility of our approach and was one of several multimodel efforts that laid the groundwork for the COVID-19 Scenario Modeling Hub, which has provided multiple rounds of real-time scenario projections for situational awareness and decision making to the Centers for Disease Control and Prevention since December 2020.

Phase transitions may explain why SARS-CoV-2 spreads so fast and why new variants are spreading faster.

The novel coronavirus SARS CoV-2 responsible for the COVID-19 pandemic and SARS CoV-1 responsible for the SARS epidemic of 2002-2003 share an ancestor yet evolved to have much different transmissibility and global impact 1. A previously developed thermodynamic model of protein conformations hypothesized that SARS CoV-2 is very close to a new thermodynamic critical point, which makes it highly infectious but also easily displaced by a spike-based vaccine because there is a tradeoff between transmissibility and robustness 2. The model identified a small cluster of four key mutations of SARS CoV-2 that predicts much stronger viral attachment and viral spreading compared to SARS CoV-1. Here we apply the model to the SARS-CoV-2 variants Alpha (B.1.1.7), Beta (B.1.351), Gamma (P.1) and Delta (B.1.617.2)3 and predict, using no free parameters, how the new mutations will not diminish the effectiveness of current spike based vaccines and may even further enhance infectiousness by augmenting the binding ability of the virus.

Divergent COVID-19 Disease Trajectories Predicted by a DAMP-Centered Immune Network Model.

COVID-19 presentations range from mild to moderate through severe disease but also manifest with persistent illness or viral recrudescence. We hypothesized that the spectrum of COVID-19 disease manifestations was a consequence of SARS-CoV-2-mediated delay in the pathogen-associated molecular pattern (PAMP) response, including dampened type I interferon signaling, thereby shifting the balance of the immune response to be dominated by damage-associated molecular pattern (DAMP) signaling. To test the hypothesis, we constructed a parsimonious mechanistic mathematical model. After calibration of the model for initial viral load and then by varying a few key parameters, we show that the core model generates four distinct viral load, immune response and associated disease trajectories termed "patient archetypes", whose temporal dynamics are reflected in clinical data from hospitalized COVID-19 patients. The model also accounts for responses to corticosteroid therapy and predicts that vaccine-induced neutralizing antibodies and cellular memory will be protective, including from severe COVID-19 disease. This generalizable modeling framework could be used to analyze protective and pathogenic immune responses to diverse viral infections.

Training Spiking Neural Networks in the Strong Coupling Regime.

Recurrent neural networks trained to perform complex tasks can provide insight into the dynamic mechanism that underlies computations performed by cortical circuits. However, due to a large number of unconstrained synaptic connections, the recurrent connectivity that emerges from network training may not be biologically plausible. Therefore, it remains unknown if and how biological neural circuits implement dynamic mechanisms proposed by the models. To narrow this gap, we developed a training scheme that, in addition to achieving learning goals, respects the structural and dynamic properties of a standard cortical circuit model: strongly coupled excitatory-inhibitory spiking neural networks. By preserving the strong mean excitatory and inhibitory coupling of initial networks, we found that most of trained synapses obeyed Dale's law without additional constraints, exhibited large trial-to-trial spiking variability, and operated in inhibition-stabilized regime. We derived analytical estimates on how training and network parameters constrained the changes in mean synaptic strength during training. Our results demonstrate that training recurrent neural networks subject to strong coupling constraints can result in connectivity structure and dynamic regime relevant to cortical circuits.

Pupal behavior emerges from unstructured muscle activity in response to neuromodulation in Drosophila.

Identifying neural substrates of behavior requires defining actions in terms that map onto brain activity. Brain and muscle activity naturally correlate via the output of motor neurons, but apart from simple movements it has been difficult to define behavior in terms of muscle contractions. By mapping the musculature of the pupal fruit fly and comprehensively imaging muscle activation at single-cell resolution, we here describe a multiphasic behavioral sequence in Drosophila. Our characterization identifies a previously undescribed behavioral phase and permits extraction of major movements by a convolutional neural network. We deconstruct movements into a syllabary of co-active muscles and identify specific syllables that are sensitive to neuromodulatory manipulations. We find that muscle activity shows considerable variability, with sequential increases in stereotypy dependent upon neuromodulation. Our work provides a platform for studying whole-animal behavior, quantifying its variability across multiple spatiotemporal scales, and analyzing its neuromodulatory regulation at cellular resolution.

How do we find out how the brain works? One way is to use imaging techniques to visualise an animal's brain in action as it performs simple behaviours: as the animal moves, parts of its brain light up under the microscope. For laboratory animals like fruit flies, which have relatively small brains, this lets us observe their brain activity right down to the level of individual brain cells. The brain directs movements via collective activity of the body's muscles. Our ability to track the activity of individual muscles is, however, more limited than our ability to observe single brain cells: even modern imaging technology still cannot monitor the activity of all the muscle cells in an animal's body as it moves about. Yet this is precisely the information that scientists need to fully understand how the brain generates behaviour. Fruit flies perform specific behaviours at certain stages of their life cycle. When the fly pupa begins to metamorphose into an adult insect, it performs a fixed sequence of movements involving a set number of muscles, which is called the pupal ecdysis sequence. This initial movement sequence and the rest of metamorphosis both occur within the confines of the pupal case, which is a small, hardened shell surrounding the whole animal. Elliott et al. set out to determine if the fruit fly pupa's ecdysis sequence could be used as a kind of model, to describe a simple behaviour at the level of individual muscles. Imaging experiments used fly pupae that were genetically engineered to produce an activity-dependent fluorescent protein in their muscle cells. Pupal cases were treated with a chemical to make them transparent, allowing easy observation of their visually 'labelled' muscles. This yielded a near-complete record of muscle activity during metamorphosis. Initially, individual muscles became active in small groups. The groups then synchronised with each other over the different regions of the pupa's body to form distinct movements, much as syllables join to form words. This synchronisation was key to progression through metamorphosis and was co-ordinated at each step by specialised nerve cells that produce or respond to specific hormones. These results reveal how the brain might direct muscle activity to produce movement patterns. In the future, Elliott et al. hope to compare data on muscle activity with comprehensive records of brain cell activity, to shed new light on how the brain, muscles, and other factors work together to control behaviour.

Dynamic imaging of nascent RNA reveals general principles of transcription dynamics and stochastic splice site selection.

The activities of RNA polymerase and the spliceosome are responsible for the heterogeneity in the abundance and isoform composition of mRNA in human cells. However, the dynamics of these megadalton enzymatic complexes working in concert on endogenous genes have not been described. Here, we establish a quasi-genome-scale platform for observing synthesis and processing kinetics of single nascent RNA molecules in real time. We find that all observed genes show transcriptional bursting. We also observe large kinetic variation in intron removal for single introns in single cells, which is inconsistent with deterministic splice site selection. Transcriptome-wide footprinting of the U2AF complex, nascent RNA profiling, long-read sequencing, and lariat sequencing further reveal widespread stochastic recursive splicing within introns. We propose and validate a unified theoretical model to explain the general features of transcription and pervasive stochastic splice site selection.

A Quantitative Evaluation of COVID-19 Epidemiological Models.

Quantifying how accurate epidemiological models of COVID-19 forecast the number of future cases and deaths can help frame how to incorporate mathematical models to inform public health decisions. Here we analyze and score the predictive ability of publicly available COVID-19 epidemiological models on the COVID-19 Forecast Hub. Our score uses the posted forecast cumulative distributions to compute the log-likelihood for held-out COVID-19 positive cases and deaths. Scores are updated continuously as new data become available, and model performance is tracked over time. We use model scores to construct ensemble models based on past performance. Our publicly available quantitative framework may aid in improving modeling frameworks, and assist policy makers in selecting modeling paradigms to balance the delicate trade-offs between the economy and public health.

Genomic analysis of diet composition finds novel loci and associations with health and lifestyle.

We conducted genome-wide association studies (GWAS) of relative intake from the macronutrients fat, protein, carbohydrates, and sugar in over 235,000 individuals of European ancestries. We identified 21 unique, approximately independent lead SNPs. Fourteen lead SNPs are uniquely associated with one macronutrient at genome-wide significance (P < 5 × 10-8), while five of the 21 lead SNPs reach suggestive significance (P < 1 × 10-5) for at least one other macronutrient. While the phenotypes are genetically correlated, each phenotype carries a partially unique genetic architecture. Relative protein intake exhibits the strongest relationships with poor health, including positive genetic associations with obesity, type 2 diabetes, and heart disease (rg ≈ 0.15-0.5). In contrast, relative carbohydrate and sugar intake have negative genetic correlations with waist circumference, waist-hip ratio, and neighborhood deprivation (|rg| ≈ 0.1-0.3) and positive genetic correlations with physical activity (rg ≈ 0.1 and 0.2). Relative fat intake has no consistent pattern of genetic correlations with poor health but has a negative genetic correlation with educational attainment (rg ≈-0.1). Although our analyses do not allow us to draw causal conclusions, we find no evidence of negative health consequences associated with relative carbohydrate, sugar, or fat intake. However, our results are consistent with the hypothesis that relative protein intake plays a role in the etiology of metabolic dysfunction.

COVID-19 reopening strategies at the county level in the face of uncertainty: Multiple Models for Outbreak Decision Support.

Policymakers make decisions about COVID-19 management in the face of considerable uncertainty. We convened multiple modeling teams to evaluate reopening strategies for a mid-sized county in the United States, in a novel process designed to fully express scientific uncertainty while reducing linguistic uncertainty and cognitive biases. For the scenarios considered, the consensus from 17 distinct models was that a second outbreak will occur within 6 months of reopening, unless schools and non-essential workplaces remain closed. Up to half the population could be infected with full workplace reopening; non-essential business closures reduced median cumulative infections by 82%. Intermediate reopening interventions identified no win-win situations; there was a trade-off between public health outcomes and duration of workplace closures. Aggregate results captured twice the uncertainty of individual models, providing a more complete expression of risk for decision-making purposes.

Dissecting transcriptional amplification by MYC.

Supraphysiological MYC levels are oncogenic. Originally considered a typical transcription factor recruited to E-boxes (CACGTG), another theory posits MYC a global amplifier increasing output at all active promoters. Both models rest on large-scale genome-wide "-omics'. Because the assumptions, statistical parameter and model choice dictates the '-omic' results, whether MYC is a general or specific transcription factor remains controversial. Therefore, an orthogonal series of experiments interrogated MYC's effect on the expression of synthetic reporters. Dose-dependently, MYC increased output at minimal promoters with or without an E-box. Driving minimal promoters with exogenous (glucocorticoid receptor) or synthetic transcription factors made expression more MYC-responsive, effectively increasing MYC-amplifier gain. Mutations of conserved MYC-Box regions I and II impaired amplification, whereas MYC-box III mutations delivered higher reporter output indicating that MBIII limits over-amplification. Kinetic theory and experiments indicate that MYC activates at least two steps in the transcription-cycle to explain the non-linear amplification of transcription that is essential for global, supraphysiological transcription in cancer.

A mathematical model for persistent post-CSD vasoconstriction.

Cortical spreading depression (CSD) is the propagation of a relatively slow wave in cortical brain tissue that is linked to a number of pathological conditions such as stroke and migraine. Most of the existing literature investigates the dynamics of short term phenomena such as the depolarization and repolarization of membrane potentials or large ion shifts. Here, we focus on the clinically-relevant hour-long state of neurovascular malfunction in the wake of CSDs. This dysfunctional state involves widespread vasoconstriction and a general disruption of neurovascular coupling. We demonstrate, using a mathematical model, that dissolution of calcium that has aggregated within the mitochondria of vascular smooth muscle cells can drive an hour-long disruption. We model the rate of calcium clearance as well as the dynamical implications on overall blood flow. Based on reaction stoichiometry, we quantify a possible impact of calcium phosphate dissolution on the maintenance of F0F1-ATP synthase activity.

Global prediction of unreported SARS-CoV2 infection from observed COVID-19 cases.

Estimation of infectiousness and fatality of the SARS-CoV-2 virus in the COVID-19 global pandemic is complicated by ascertainment bias resulting from incomplete and non-representative samples of infected individuals. We developed a strategy for overcoming this bias to obtain more plausible estimates of the true values of key epidemiological variables. We fit mechanistic Bayesian latent-variable SIR models to confirmed COVID-19 cases, deaths, and recoveries, for all regions (countries and US states) independently. Bayesian averaging over models, we find that the raw infection incidence rate underestimates the true rate by a factor, the case ascertainment ratio CARt that depends upon region and time. At the regional onset of COVID-19, the predicted global median was 13 infections unreported for each case confirmed (CARt = 0.07 C.I. (0.02, 0.4)). As the infection spread, the median CARt rose to 9 unreported cases for every one diagnosed as of April 15, 2020 (CARt = 0.1 C.I. (0.02, 0.5)). We also estimate that the median global initial reproduction number R0 is 3.3 (C.I (1.5, 8.3)) and the total infection fatality rate near the onset is 0.17% (C.I. (0.05%, 0.9%)). However the time-dependent reproduction number Rt and infection fatality rate as of April 15 were 1.2 (C.I. (0.6, 2.5)) and 0.8% (C.I. (0.2%,4%)), respectively. We find that there is great variability between country- and state-level values. Our estimates are consistent with recent serological estimates of cumulative infections for the state of New York, but inconsistent with claims that very large fractions of the population have already been infected in most other regions. For most regions, our estimates imply a great deal of uncertainty about the current state and trajectory of the epidemic.

Before and beyond the Wilson-Cowan equations.

The Wilson-Cowan equations represent a landmark in the history of computational neuroscience. Along with the insights Wilson and Cowan offered for neuroscience, they crystallized an approach to modeling neural dynamics and brain function. Although their iconic equations are used in various guises today, the ideas that led to their formulation and the relationship to other approaches are not well known. Here, we give a little context to some of the biological and theoretical concepts that lead to the Wilson-Cowan equations and discuss how to extend beyond them.

Dynamical modeling of multi-scale variability in neuronal competition.

Variability is observed at multiple-scales in the brain and ubiquitous in perception. However, the nature of perceptual variability is an open question. We focus on variability during perceptual rivalry, a form of neuronal competition. Rivalry provides a window into neural processing since activity in many brain areas is correlated to the alternating perception rather than a constant ambiguous stimulus. It exhibits robust properties at multiple scales including conscious awareness and neuron dynamics. The prevalent theory for spiking variability is called the balanced state; whereas, the source of perceptual variability is unknown. Here we show that a single biophysical circuit model, satisfying certain mutual inhibition architectures, can explain spiking and perceptual variability during rivalry. These models adhere to a broad set of strict experimental constraints at multiple scales. As we show, the models predict how spiking and perceptual variability changes with stimulus conditions.

Unintentional error in formula preparation and its simulated impact on infant weight and adiposity.

BACKGROUND: Accelerated weight gain in infancy is a public health issue and is likely due to feeding behaviours. OBJECTIVES: To test the accuracy of individuals to dispense infant formula as compared with recommended serving sizes and to estimate the effect of dispensing inaccuracy on infant growth. METHODS: Fifty-three adults dispensed infant formula powder for three servings of 2, 4, 6, and 8 fl oz bottles, in random order. The weight of dispensed infant formula powder was compared with the recommended serving size weight on the nutrition label. A novel mathematical model was used to estimate the impact of formula dispensing on infant weight and adiposity. RESULTS: Nineteen percent of bottles (20 of 636) prepared contained the recommended amount of infant formula powder. Three percent were underdispensed, and 78% of bottles were overdispensed, resulting in 11% additional infant formula powder. Mathematical modelling feeding 11% above energy requirements exclusively for 6 months for male and female infants suggested infants at the 50th percentile for weight at birth would reach the 75th percentile with increased adiposity by 6 months. CONCLUSIONS: Inaccurate measurement of infant formula powder and overdispensing, which is highly prevalent, specifically, may contribute to rapid weight gain and increased adiposity in formula-fed infants.