Modeling the Control of COVID via Non-Pharmaceutical Intervention

The COVID-19 outbreak gave rise to an unprecedented production of models and studies aimed at understanding the pandemic, predicting its evolution and designing measures to reduce its spread.The aim of this project is to show how a simple SIR model was used to make quick predictions for New Jersey in early March 2020 and call for action based on data from China and Italy. Now different viruses manifest with different characteristics and public response to these characteristics can be drastically different. Therefore A more refined model, which accounts for the parameters social distancing, testing, contact tracing and quarantining, is then proposed to identify containment measures to minimize the economic cost of the pandemic.

This model was programmed using AMPL (a mathematical programming language) in which we use optimization techniques and data from throughout New Jersey, split into three regions, to minimize the economic costs of the aforementioned parameters. For visualization and plotting we use Matlab to plot our results.

Funding:  This work is supported by the NSF CMMI project 2033580 "Managing pandemic by managing mobility" in collaboration with Cornell University and Vanderbilt University, and the support of the Joseph and Loretta Lopez Chair endowment.

Students: Ryan Weightman, Sean McQuade

Optimized COVID Vaccination Plan

The COVID-19 outbreak gave rise to an unprecedented production of models and studies aimed at understanding the pandemic, predicting its evolution and designing measures to reduce its spread. There was an especially strong focus on how to accurately model vaccination of a certain population in a compartmental SIR model. We build a SIR model with vaccination compartments and exposed compartment transforming it into an SVEIR model. We then split the population into major age groups to better capture the varying effect of the virus on specific populations. Lastly, we optimize vaccine schedule to minimize deaths amongst the population. All of this is done in Python using an optimation package.

Funding:  This work is supported by the NSF CMMI project 2033580 "Managing pandemic by managing mobility" in collaboration with Cornell University and Vanderbilt University, and the support of the Joseph and Loretta Lopez Chair endowment.

Students: Ryan Weightman, Sean McQuade

Modeling metabolic systems via Linear-In-Flux-Expressions (LIFE)

We have designed a framework for modeling systems of biochemical reactions. Our research addresses the foundation of modeling complex reactions (between three or more molecules) and the capability of a drug to inhibit or enhance fluxes in the system. We introduce the concept of metabolic graphs, a generalization of hypergraphs having specialized features common to metabolic networks; these features are visualizations of the framework that corresponds to complex reaction dynamics and drug inhibition or enhancement.

LIFE utilizes the flux space at equilibrium to reduce the total number model parameters. This allows one to impose the equilibrium structure of the model while exploring the sensitivity of the network to perturbations, such as drug treatment or illness. We have developed simulations to implement these tools, allowing us to analyze metabolic response to drug treatment. This is valuable to quantitative systems pharmacology, since combination treatment is time consuming and expensive to perform many experiments corresponding to the many potential combinations of drugs.

Funding:  This work is supported by the Joseph and Loretta Lopez Chair endowment.

Students: Sean McQuade, Christopher Denaro, Heba Yousef

Linear-In-Flux-Expressions (LIFE) Implementation via Machine Learning Techniques

We show that deep learning models, and especially architectures like the Transformer, originally intended for natural language, can be trained on randomly generated datasets to predict to very high accuracy both the qualitative and quantitative features of metabolic networks. Using standard mathematical techniques, we create large sets (40 million cases) of random networks that can be used to train our models. These trained models can predict network equilibrium on random graphs in more than 99% of cases. They can also generalize to graphs with different structure than those encountered at training. Finally, they can predict almost perfectly the equilibria of a small set of known biological networks. Our approach is both very economical in experimental data and uses only small and shallow deep-learning model, far from the large architectures commonly used in machine translation. Such results pave the way for larger use of deep learning models for problems related to biological networks in key areas such as quantitative systems pharmacology, systems biology, and synthetic biology.

Funding:  This work is supported by the Joseph and Loretta Lopez Chair endowment.

Students: Christopher Denaro, Sean McQuade, Heba Yousef

CIRCLES Consortium

We aim to improve the future of transportation and advance the convergence of artificial intelligence, simulation, traffic engineering, and vehicle technology in the context of mixed human-autonomous traffic.

The Piccoli lab develops control algorithms based on the mean field limit of microscopic traffic models. This leads to models of ODEs and PDEs that are the basis for optimal control problems designed to minimize the energy usage.

Funding:  This research is based upon work supported by the U.S. Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE) under the Vehicle Technologies Office award number CID DE-EE0008872.

Students: Sean McQuade