Parallel isotope differential modeling for instationary 13C fluxomics at the genome scale

Zhengdong Zhang; Zhentao Liu; Yafei Meng; Zhen Chen; Jiayu Han; Yimin Wei; Tie Shen; Yin Yi; Xiaoyao Xie

doi:10.1186/s13068-020-01737-5

Parallel isotope differential modeling for instationary 13C fluxomics at the genome scale

Biotechnol Biofuels. 2020 Jun 8:13:103. doi: 10.1186/s13068-020-01737-5. eCollection 2020.

Authors

Zhengdong Zhang^#^{1

2}, Zhentao Liu^#³, Yafei Meng¹, Zhen Chen⁴, Jiayu Han⁴, Yimin Wei⁵, Tie Shen², Yin Yi⁶, Xiaoyao Xie²

Affiliations

¹ College of Mathematics and Information Science, Guiyang University, Guiyang, Guizhou China.
² Key Laboratory of Information and Computing Science Guizhou Province, Guizhou Normal University, Guiyang, Guizhou China.
³ College of Computer Science and Technology, Guizhou University, Guiyang, Guizhou China.
⁴ School of Mathematics and Sciences, Guizhou Normal University, Guiyang, Guizhou China.
⁵ School of Mathematics Sciences and Key Laboratory of Mathematics for Nonlinear Sciences, Fudan University, Shanghai, China.
⁶ College of Life Science, Guizhou Normal University, Guiyang, Guizhou China.

^# Contributed equally.

Abstract

Background: A precise map of the metabolic fluxome, the closest surrogate to the physiological phenotype, is becoming progressively more important in the metabolic engineering of photosynthetic organisms for biofuel and biomass production. For photosynthetic organisms, the state-of-the-art method for this purpose is instationary 13C fluxomics, which has arisen as a sibling of transcriptomics or proteomics. Instationary 13C data processing requires solving high-dimensional nonlinear differential equations and leads to large computational and time costs when its scope is expanded to a genome-scale metabolic network.

Result: Here, we present a parallelized method to model instationary 13C labeling data. The elementary metabolite unit (EMU) framework is reorganized to allow treating individual mass isotopomers and breaking up of their networks into strongly connected components (SCCs). A variable domain parallel algorithm is introduced to process ordinary differential equations in a parallel way. 15-fold acceleration is achieved for constant-step-size modeling and ~ fivefold acceleration for adaptive-step-size modeling.

Conclusion: This algorithm is universally applicable to isotope granules such as EMUs and cumomers and can substantially accelerate instationary 13C fluxomics modeling. It thus has great potential to be widely adopted in any instationary 13C fluxomics modeling.

Keywords: 13C fluxomics; Genome-scale metabolic flux analysis; Instationary metabolic flux analysis; Mass isotopomer network; Parallel differential equations modeling.