Enhanced SARS-CoV-2 case prediction using public health data and machine learning models.

Objectives: The goal of this study is to propose and test a scalable framework for machine learning (ML) algorithms to predict near-term severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) cases by incorporating and evaluating the impact of real-time dynamic public health data. Materials and Methods: Data used in this study include patient-level results, procurement, and location information of all SARS-CoV-2 tests reported in West Virginia as part of their mandatory reporting system from January 2021 to March 2022. We propose a method for incorporating and comparing widely available public health metrics inside of a ML framework, specifically a long-short-term memory network, to forecast SARS-CoV-2 cases across various feature sets. Results: Our approach provides better prediction of localized case counts and indicates the impact of the dynamic elements of the pandemic on predictions, such as the influence of the mixture of viral variants in the population and variable testing and vaccination rates during various eras of the pandemic. Discussion: Utilizing real-time public health metrics, including estimated Rt from multiple SARS-CoV-2 variants, vaccination rates, and testing information, provided a significant increase in the accuracy of the model during the Omicron and Delta period, thus providing more precise forecasting of daily case counts at the county level. This work provides insights on the influence of various features on predictive performance in rural and non-rural areas. Conclusion: Our proposed framework incorporates available public health metrics with operational data on the impact of testing, vaccination, and current viral variant mixtures in the population to provide a foundation for combining dynamic public health metrics and ML models to deliver forecasting and insights in healthcare domains. It also shows the importance of developing and deploying ML frameworks in rural settings.

SARS-CoV-2 variant identification using a genome tiling array and genotyping probes.

With over 5.5 million deaths worldwide attributed to the respiratory disease COVID-19 caused by the novel coronavirus SARS-CoV-2, it is essential that continued efforts be made to track the evolution and spread of the virus globally. The authors previously presented a rapid and cost-effective method to sequence the entire SARS-CoV-2 genome with 95% coverage and 99.9% accuracy. This method is advantageous for identifying and tracking variants in the SARS-CoV-2 genome compared with traditional short-read sequencing methods which can be time-consuming and costly. Herein, the addition of genotyping probes to a DNA chip that targets known SARS-CoV-2 variants is presented. The incorporation of genotyping probe sets along with the advent of a moving average filter improved the sequencing coverage and accuracy of the SARS-CoV-2 genome.

Throughout the COVID-19 pandemic the virus known as SARS-CoV-2 has continued to mutate and evolve. It is imperative to continue to track these mutations and where the virus has traveled to best inform healthcare practices and global strategies to combat the virus. The authors previously developed a method to investigate 95% of this viral genome with 99.9% accuracy that was more cost-effective and less time-consuming than previous methods. In this work, specific markers were added to the technology to allow tracking of mutations in the virus that have already been documented. In doing so, the accuracy and how much of the viral genome can be sequenced was improved.

Spatial Stochastic Model of the Pre-B Cell Receptor.

Survival and proliferation of immature B lymphocytes requires expression and tonic signaling of the pre-B cell receptor (pre-BCR). This low level, ligand-independent signaling is likely achieved through frequent, but short-lived, homo interactions. Tonic signaling is also central in the pathology of precursor B acute lymphoblastic leukemia (B-ALL). In order to understand how repeated, transient events can lead to sustained signaling and to assess the impact of receptor accumulation induced by the membrane landscape, we developed a spatial stochastic model of receptor aggregation and downstream signaling events. Our rule- and agent-based model builds on previous mature BCR signaling models and incorporates novel parameters derived from single particle tracking of pre-BCR on surfaces of two different B-ALL cell lines, 697 and Nalm6. Live cell tracking of receptors on the two cell lines revealed characteristic differences in their dimer dissociation rates and diffusion coefficients. We report here that these differences affect pre-BCR aggregation and consequent signal initiation events. Receptors on Nalm6 cells, which have a lower off-rate and lower diffusion coefficient, more frequently form higher order oligomers than pre-BCR on 697 cells, resulting in higher levels of downstream phosphorylation in the Nalm6 cell line.

Modeling the Cluster Size Distribution of Vascular Endothelial Growth Factor (VEGF) Receptors.

We previously developed a method of defining receptor clusters in the membrane based on mutual distance and applied it to a set of transmission microscopy images of vascular endothelial growth factor receptors. An optimal length parameter was identified, resulting in cluster identification and a procedure that assigned a geometric shape to each cluster. We showed that the observed particle distribution results were consistent with the random placement of receptors within the clusters and, to a lesser extent, the random placement of the clusters on the cell membrane. Here, we develop and validate a stochastic model of clustering, based on a hypothesis of preexisting domains that have a high affinity for receptors. The proximate objective is to clarify the mechanism behind cluster formation and to estimate the effect on signaling. Receptor-enriched domains may significantly impact signaling pathways that rely on ligand-induced dimerization of receptors. We define a simple statistical model, based on the preexisting domain hypothesis, to predict the probability distribution of cluster sizes. The process yielded sets of parameter values that can readily be used in dynamical calculations as the estimates of the quantitative characteristics of the clustering domains.

Predicting increases in COVID-19 incidence to identify locations for targeted testing in West Virginia: A machine learning enhanced approach.

During the COVID-19 pandemic, West Virginia developed an aggressive SARS-CoV-2 testing strategy which included utilizing pop-up mobile testing in locations anticipated to have near-term increases in SARS-CoV-2 infections. This study describes and compares two methods for predicting near-term SARS-CoV-2 incidence in West Virginia counties. The first method, Rt Only, is solely based on producing forecasts for each county using the daily instantaneous reproductive numbers, Rt. The second method, ML+Rt, is a machine learning approach that uses a Long Short-Term Memory network to predict the near-term number of cases for each county using epidemiological statistics such as Rt, county population information, and time series trends including information on major holidays, as well as leveraging statewide COVID-19 trends across counties and county population size. Both approaches used daily county-level SARS-CoV-2 incidence data provided by the West Virginia Department Health and Human Resources beginning April 2020. The methods are compared on the accuracy of near-term SARS-CoV-2 increases predictions by county over 17 weeks from January 1, 2021- April 30, 2021. Both methods performed well (correlation between forecasted number of cases and the actual number of cases week over week is 0.872 for the ML+Rt method and 0.867 for the Rt Only method) but differ in performance at various time points. Over the 17-week assessment period, the ML+Rt method outperforms the Rt Only method in identifying larger spikes. Results show that both methods perform adequately in both rural and non-rural predictions. Finally, a detailed discussion on practical issues regarding implementing forecasting models for public health action based on Rt is provided, and the potential for further development of machine learning methods that are enhanced by Rt.

Predicting increases in COVID-19 incidence to identify locations for targeted testing in West Virginia: A machine learning enhanced approach.

During the COVID-19 pandemic, West Virginia developed an aggressive SARS-CoV-2 testing strategy which included utilizing pop-up mobile testing in locations anticipated to have near-term increases in SARS-CXoV-2 infections. In this study, we describe and compare two methods for predicting near-term SARS-CoV-2 incidence in West Virginia counties. The first method, R t Only, is solely based on producing forecasts for each county using the daily instantaneous reproductive numbers, R t. The second method, ML+ R t , is a machine learning approach that uses a Long Short-Term Memory network to predict the near-term number of cases for each county using epidemiological statistics such as Rt, county population information, and time series trends including information on major holidays, as well as leveraging statewide COVID-19 trends across counties and county population size. Both approaches used daily county-level SARS-CoV-2 incidence data provided by the West Virginia Department Health and Human Resources beginning April 2020. The methods are compared on the accuracy of near-term SARS-CoV-2 increases predictions by county over 17 weeks from January 1, 2021-April 30, 2021. Both methods performed well (correlation between forecasted number of cases and the actual number of cases week over week is 0.872 for the ML+R t method and 0.867 for the R t Only method) but differ in performance at various time points. Over the 17-week assessment period, the ML+R t method outperforms the R t Only method in identifying larger spikes. We also find that both methods perform adequately in both rural and non-rural predictions. Finally, we provide a detailed discussion on practical issues regarding implementing forecasting models for public health action based on R t , and the potential for further development of machine learning methods that are enhanced by R t.

SARS-CoV-2 Variant Identification Using a Genome Tiling Array and Genotyping Probes.

With over three million deaths worldwide attributed to the respiratory disease COVID-19 caused by the novel coronavirus SARS-CoV-2, it is essential that continued efforts be made to track the evolution and spread of the virus globally. We previously presented a rapid and cost-effective method to sequence the entire SARS-CoV-2 genome with 95% coverage and 99.9% accuracy. This method is advantageous for identifying and tracking variants in the SARS-CoV-2 genome when compared to traditional short read sequencing methods which can be time consuming and costly. Herein we present the addition of genotyping probes to our DNA chip which target known SARS-CoV-2 variants. The incorporation of the genotyping probe sets along with the advent of a moving average filter have improved our sequencing coverage and accuracy of the SARS-CoV-2 genome.

Highly Accurate Chip-Based Resequencing of SARS-CoV-2 Clinical Samples.

SARS-CoV-2 has infected over 128 million people worldwide, and until a vaccine is developed and widely disseminated, vigilant testing and contact tracing are the most effective ways to slow the spread of COVID-19. Typical clinical testing only confirms the presence or absence of the virus, but rather, a simple and rapid testing procedure that sequences the entire genome would be impactful and allow for tracing the spread of the virus and variants, as well as the appearance of new variants. However, traditional short read sequencing methods are time consuming and expensive. Herein, we describe a tiled genome array that we developed for rapid and inexpensive full viral genome resequencing, and we have applied our SARS-CoV-2-specific genome tiling array to rapidly and accurately resequence the viral genome from eight clinical samples. We have resequenced eight samples acquired from patients in Wyoming that tested positive for SARS-CoV-2. We were ultimately able to sequence over 95% of the genome of each sample with greater than 99.9% average accuracy.

Effect of Spatial Inhomogeneities on the Membrane Surface on Receptor Dimerization and Signal Initiation.

Important signal transduction pathways originate on the plasma membrane, where microdomains may transiently entrap diffusing receptors. This results in a non-random distribution of receptors even in the resting state, which can be visualized as "clusters" by high resolution imaging methods. Here, we explore how spatial in-homogeneities in the plasma membrane might influence the dimerization and phosphorylation status of ErbB2 and ErbB3, two receptor tyrosine kinases that preferentially heterodimerize and are often co-expressed in cancer. This theoretical study is based upon spatial stochastic simulations of the two-dimensional membrane landscape, where variables include differential distributions and overlap of transient confinement zones ("domains") for the two receptor species. The in silico model is parameterized and validated using data from single particle tracking experiments. We report key differences in signaling output based on the degree of overlap between domains and the relative retention of receptors in such domains, expressed as escape probability. Results predict that a high overlap of domains, which favors transient co-confinement of both receptor species, will enhance the rate of hetero-interactions. Where domains do not overlap, simulations confirm expectations that homo-interactions are favored. Since ErbB3 is uniquely dependent on ErbB2 interactions for activation of its catalytic activity, variations in domain overlap or escape probability markedly alter the predicted patterns and time course of ErbB3 and ErbB2 phosphorylation. Taken together, these results implicate membrane domain organization as an important modulator of signal initiation, motivating the design of novel experimental approaches to measure these important parameters across a wider range of receptor systems.

Orchestration of ErbB3 signaling through heterointeractions and homointeractions.

Members of the ErbB family of receptor tyrosine kinases are capable of both homointeractions and heterointeractions. Because each receptor has a unique set of binding sites for downstream signaling partners and differential catalytic activity, subtle shifts in their combinatorial interplay may have a large effect on signaling outcomes. The overexpression and mutation of ErbB family members are common in numerous human cancers and shift the balance of activation within the signaling network. Here we report the development of a spatial stochastic model that addresses the dynamics of ErbB3 homodimerization and heterodimerization with ErbB2. The model is based on experimental measures for diffusion, dimer off-rates, kinase activity, and dephosphorylation. We also report computational analysis of ErbB3 mutations, generating the prediction that activating mutations in the intracellular and extracellular domains may be subdivided into classes with distinct underlying mechanisms. We show experimental evidence for an ErbB3 gain-of-function point mutation located in the C-lobe asymmetric dimerization interface, which shows enhanced phosphorylation at low ligand dose associated with increased kinase activity.

Analytical solution of steady-state equations for chemical reaction networks with bilinear rate laws.

True steady states are a rare occurrence in living organisms, yet their knowledge is essential for quasi-steady-state approximations, multistability analysis, and other important tools in the investigation of chemical reaction networks (CRN) used to describe molecular processes on the cellular level. Here, we present an approach that can provide closed form steady-state solutions to complex systems, resulting from CRN with binary reactions and mass-action rate laws. We map the nonlinear algebraic problem of finding steady states onto a linear problem in a higher-dimensional space. We show that the linearized version of the steady-state equations obeys the linear conservation laws of the original CRN. We identify two classes of problems for which complete, minimally parameterized solutions may be obtained using only the machinery of linear systems and a judicious choice of the variables used as free parameters. We exemplify our method, providing explicit formulae, on CRN describing signal initiation of two important types of RTK receptor-ligand systems, VEGF and EGF-ErbB1.

Dynamic transition states of ErbB1 phosphorylation predicted by spatial stochastic modeling.

ErbB1 overexpression is strongly linked to carcinogenesis, motivating better understanding of erbB1 dimerization and activation. Recent single-particle-tracking data have provided improved measures of dimer lifetimes and strong evidence that transient receptor coconfinement promotes repeated interactions between erbB1 monomers. Here, spatial stochastic simulations explore the potential impact of these parameters on erbB1 phosphorylation kinetics. This rule-based mathematical model incorporates structural evidence for conformational flux of the erbB1 extracellular domains, as well as asymmetrical orientation of erbB1 cytoplasmic kinase domains during dimerization. The asymmetric dimer model considers the theoretical consequences of restricted transactivation of erbB1 receptors within a dimer, where the N-lobe of one monomer docks with the C-lobe of the second monomer and triggers its catalytic activity. The dynamic nature of the erbB1 phosphorylation state is shown by monitoring activation states of individual monomers as they diffuse, bind, and rebind after ligand addition. The model reveals the complex interplay between interacting liganded and nonliganded species and the influence of their distribution and abundance within features of the membrane landscape.

The impact of high density receptor clusters on VEGF signaling.

Vascular endothelial growth factor (VEGF) signaling is involved in the process of blood vessel development and maintenance. Signaling is initiated by binding of the bivalent VEGF ligand to the membrane-bound receptors (VEGFR), which in turn stimulates receptor dimerization. Herein, we discuss experimental evidence that VEGF receptors localize in caveloae and other regions of the plasma membrane, and for other receptors, it has been shown that receptor clustering has an impact on dimerization and thus also on signaling. Overall, receptor clustering is part of a complex ecosystem of interactions and how receptor clustering impacts dimerization is not well understood. To address these questions, we have formulated the simplest possible model. We have postulated the existence of a single high affinity region in the cell membrane, which acts as a transient trap for receptors. We have defined an ODE model by introducing high- and low-density receptor variables and introduce the corresponding reactions from a realistic model of VEGF signal initiation. Finally, we use the model to investigate the relation between the degree of VEGFR concentration, ligand availability, and signaling. In conclusion, our simulation results provide a deeper understanding of the role of receptor clustering in cell signaling.

Mathematical simulation of membrane protein clustering for efficient signal transduction.

Initiation and propagation of cell signaling depend on productive interactions among signaling proteins at the plasma membrane. These diffusion-limited interactions can be influenced by features of the membrane that introduce barriers, such as cytoskeletal corrals, or microdomains that transiently confine both transmembrane receptors and membrane-tethered peripheral proteins. Membrane topographical features can lead to clustering of receptors and other membrane components, even under very dynamic conditions. This review considers the experimental and mathematical evidence that protein clustering impacts cell signaling in complex ways. Simulation approaches used to consider these stochastic processes are discussed.

Quantitative understanding of cell signaling: the importance of membrane organization.

Systems biology modeling of signal transduction pathways traditionally employs ordinary differential equations, deterministic models based on the assumptions of spatial homogeneity. However, this can be a poor approximation for certain aspects of signal transduction, especially its initial steps: the cell membrane exhibits significant spatial organization, with diffusion rates approximately two orders of magnitude slower than those in the cytosol. Thus, to unravel the complexities of signaling pathways, quantitative models must consider spatial organization as an important feature of cell signaling. Furthermore, spatial separation limits the number of molecules that can physically interact, requiring stochastic simulation methods that account for individual molecules. Herein, we discuss the need for mathematical models and experiments that appreciate the importance of spatial organization in the membrane.

Towards A Fully Automated High-Throughput Phototransfection System.

We have designed and implemented a framework for creating a fully automated high-throughput phototransfection system. Integrated image processing, laser target position calculation, and stage movements show a throughput increase of > 23X over the current manual phototransfection method while the potential for even greater throughput improvements (> 110X) is described. A software tool for automated off-line single cell morphological measurements, as well as real-time image segmentation analysis, has also been constructed and shown to be able quantify changes in the cell before and after the process, successfully characterizing them, using metrics such as cell perimeter, area, major and minor axis length, and eccentricity values.

Systematic analysis of conservation relations in Escherichia coli genome-scale metabolic network reveals novel growth media.

A biochemical species is called producible in a constraints-based metabolic model if a feasible steady-state flux configuration exists that sustains its nonzero concentration during growth. Extreme semipositive conservation relations (ESCRs) are the simplest semipositive linear combinations of species concentrations that are invariant to all metabolic flux configurations. In this article, we outline a fundamental relationship between the ESCRs of a metabolic network and the producibility of a biochemical species under a nutrient media. We exploit this relationship in an algorithm that systematically enumerates all minimal nutrient sets that render an objective species weakly producible (i.e., producible in the absence of thermodynamic constraints) through a simple traversal of ESCRs. We apply our results to a recent genome scale model of Escherichia coli metabolism, in which we traverse the 51 anhydrous ESCRs of the metabolic network to determine all 928 minimal aqueous nutrient media that render biomass weakly producible. Applying irreversibility constraints, we find 287 of these 928 nutrient sets to be thermodynamically feasible. We also find that an additional 365 of these nutrient sets are thermodynamically feasible in the presence of oxygen. Since biomass producibility is commonly used as a surrogate for growth in genome scale metabolic models, our results represent testable hypotheses of alternate growth media derived from in silico analysis of the E. coli genome scale metabolic network.

Investigating metabolite essentiality through genome-scale analysis of Escherichia coli production capabilities.

MOTIVATION: A phenotype mechanism is classically derived through the study of a set of mutants and comparison of their biochemical capabilities. One method of comparing mutant capabilities is to characterize producible and knocked out metabolites. However such an effect is difficult to manually assess, especially for a large biochemical network and a complex media. Current algorithmic approaches towards analyzing metabolic networks either do not address this specific property or are computationally infeasible on the genome-scale. RESULTS: We have developed a novel genome-scale computational approach that identifies the full set of biochemical species that are knocked out from the metabolome following a gene deletion. Results from this approach are combined with data from in vivo mutant screens to examine the essentiality of metabolite production for a phenotype. This approach can also be a useful tool for metabolic network annotation validation and refinement in newly sequenced organisms. Combining an in silico genome-scale model of Escherichia coli metabolism with in vivo survival data, we uncover possible essential roles for several cell membranes, cell walls, and quinone species. We also identify specific biomass components whose production appears to be non-essential for survival, contrary to the assumptions of previous models. AVAILABILITY: Programs are available upon request from the authors in the form of Matlab script files. SUPPLEMENTARY INFORMATION: http://www.cis.upenn.edu/biocomp/manuscripts/bioinformatics_bti245/supp-info.html.