Does HPC Offer Advantages in Metabolic Network Modeling, Simulation and Optimization?

Project Information

fortran
Project Status: Halted
Project Region: CAREERS
Submitted By: Gaurav Khanna
Project Email: alucia@uri.edu
Project Institution: University of Rhode Island
Anchor Institution: CR-University of Rhode Island

Mentors: Angelo Lucia
Students: Eason Chen

Project Description

Metabolic network simulation/optimization requires the modeling of many interconnected metabolic pathways containing large numbers of fluxes and considerable amounts of inter-pathway transport that is executed in an outer loop of the computations.
I have a current Fortran serial code for general metabolic network modeling, simulation, and optimization based on Nash Equilibrium and Monte Carlo optimization. While robust and fast, the serial code suffers from compiler-related (Lahey-Fujitsu Fortran 95) workspace limitations when using Monte Carlo optimization and eventually all aspects of the computations will become slower as the size of the networks I want to consider grows.
HPC offers the possibility of simulating each pathway individually on a separate node and MPI to synchronize the set of pathways. However, while it is unclear to me if HPC will resolve some of the current issues with the serial code and provide real advantages with regard to simulation or optimization. Nevertheless, I believe it is certainly worth pursuing in order to determine if this represents a future path for solving this class of problems.

The student will adapt the approach above for execution in an HPC environment. As a result they will get significant experience working with compiled code (Fortran), UNIX command-line, parallel and batch computing in addition to aspects of metabolic network models

Project Information

fortran
Project Status: Halted
Project Region: CAREERS
Submitted By: Gaurav Khanna
Project Email: alucia@uri.edu
Project Institution: University of Rhode Island
Anchor Institution: CR-University of Rhode Island

Mentors: Angelo Lucia
Students: Eason Chen

Project Description

Metabolic network simulation/optimization requires the modeling of many interconnected metabolic pathways containing large numbers of fluxes and considerable amounts of inter-pathway transport that is executed in an outer loop of the computations.
I have a current Fortran serial code for general metabolic network modeling, simulation, and optimization based on Nash Equilibrium and Monte Carlo optimization. While robust and fast, the serial code suffers from compiler-related (Lahey-Fujitsu Fortran 95) workspace limitations when using Monte Carlo optimization and eventually all aspects of the computations will become slower as the size of the networks I want to consider grows.
HPC offers the possibility of simulating each pathway individually on a separate node and MPI to synchronize the set of pathways. However, while it is unclear to me if HPC will resolve some of the current issues with the serial code and provide real advantages with regard to simulation or optimization. Nevertheless, I believe it is certainly worth pursuing in order to determine if this represents a future path for solving this class of problems.

The student will adapt the approach above for execution in an HPC environment. As a result they will get significant experience working with compiled code (Fortran), UNIX command-line, parallel and batch computing in addition to aspects of metabolic network models