5th Conference on Constraint-Based Reconstruction and Analysis (COBRA 2018)

COBRA.jl - Gearing up for the huge scale

Authors

Heirendt, L. - Presenter, University of Luxembourg
Arreckx, S., University of Luxembourg
Satagopam, V. P., University of Luxembourg
Schneider, R., University of Luxembourg
Thiele, I., University of Luxembourg
Fleming, R. M. T., Leiden University
Ever larger biochemical networks are being developed by the COnstraint-Based Reconstruction and Analysis (COBRA) community. Reconstructions with millions of biochemical reactions are within reach, and standard implementations of COBRA software written in MATLAB, Python, or C, hit their limit. Existing implementations are not suitable for simultaneously analyzing thousands of these large networks and simulations are months-long.

Accelerating existing and new COBRA methods bears the challenges of short development times and speeding up analyses significantly, all while assuring scalability and parallelism across multiple computing nodes and without recurring to complex message-passing interfaces. The Julia language [1] satisfies this need. It is set to become the language of choice to accelerate analyses of large (< 500k biochemical reactions) and huge-scale biochemical networks (> 500k reactions).

A step towards gearing up for high-dimensional COBRA modeling is the release of the high-level, high-performance, and open-source COBRA.jl package (git.io/COBRA.jl). DistributedFBA.jl [2], part of COBRA.jl, allows performing a flux balance analysis and many related analysis types efficiently, especially on large and huge-scale models. Most COBRA analyses are based on the COBRA Toolbox [3], a comprehensive software suite of interoperable methods written in MATLAB, which has found widespread applications in biology, biomedicine, and biotechnology. PALM.jl allows launching analyses across several computing nodes simultaneously, and this for thousands of models at once.

Through the open-source high-performance COBRA.jl package, the analysis capabilities of the COBRA community are lifted to another level, as reconstruction and analysis of large and huge-scale models are made possible and accelerated.

[1] Bezanson et al., Julia: A Fresh Approach to Numerical Computing, SIAM Review, 59(1):65–98, 2017.

[2] Heirendt et al., DistributedFBA.jl: high-level, high-performance flux balance analysis in Julia, Bioinformatics, 33(9):1421–1423, 2017.

[3] Heirendt & Arreckx et al., Creation and analysis of biochemical constraint-based models: the COBRA Toolbox v3.0, Nature Protocols (accepted), 2018.