Abstract
BackgroundGenome-scale metabolic models are increasingly employed to predict the phenotype of various biological systems pertaining to healthcare and bioengineering. To characterize the full metabolic spectrum of such systems, Fast Flux Variability Analysis (FFVA) is commonly used in parallel with static load balancing. This approach assigns to each core an equal number of biochemical reactions without consideration of their solution complexity.ResultsHere, we present Very Fast Flux Variability Analysis (VFFVA) as a parallel implementation that dynamically balances the computation load between the cores in runtime which guarantees equal convergence time between them. VFFVA allowed to gain a threefold speedup factor with coupled models and up to 100 with ill-conditioned models along with a 14-fold decrease in memory usage.ConclusionsVFFVA exploits the parallel capabilities of modern machines to enable biological insights through optimizing systems biology modeling. VFFVA is available in C, MATLAB, and Python at https://github.com/marouenbg/VFFVA.
Highlights
Genome-scale metabolic models are increasingly employed to predict the phenotype of various biological systems pertaining to healthcare and bioengineering
The Open Multi-Processing (OpenMP)/Message Passing Interface (MPI) hybrid implementation of Very Fast Flux Variability Analysis (VFFVA) allowed to gain a significant speedup over the static load balancing approach
The run times of VFFVA were compared to Fast Flux Variability Analysis (FFVA) at different settings followed by a comparison of the different strategies of load balancing with respect to their impact on the run time per worker
Summary
The OpenMP/MPI hybrid implementation of VFFVA allowed to gain a significant speedup over the static load balancing approach. VFFVA provides the user with a similar level of flexibility as it supports both types of systems while guaranteeing the same numerical results as FVA in double precision (Figure S1) It allows accessing advanced features of OpenMP and MPI such as dynamic load balancing. With dynamic load balancing, VFFVA allowed to update the assigned chunks of iterations to every worker dynamically, which guarantees an equal distribution of the load. The computation of the solution space for WBM Homo sapiens metabolic model [2] (Fig. 4a) had a twofold speedup with 16 threads using a chunk size of 50 (806 mn) compared to FFVA (1611mn). The uniform sampling of the solution space is a time-consuming process that starts with the generation of warmup points to determine the initial starting points for sampling This step is formulated to FVA and could be accelerated using dynamic load balancing. The dynamic load balancing strategy allows efficient parallel solving of metabolic models through accelerating the computation of FVA and the fast preprocessing of sampling points thereby enabling the modeler to tackle large-scale metabolic models
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.