ORCA: a COBRA toolbox extension for model-driven discovery and analysis

BIOINFORMATICS APPLICATIONS NOTE Systems biology Vol. 30 no. 4 2014, pages 584–585 doi:10.1093/bioinformatics/btt723 Advance Access publication December 13, 2013 ORCA: a COBRA toolbox extension for model-driven discovery and analysis Longfei Mao* and Wynand S. Verwoerd Department of Molecular Biosciences, Centre for Advanced Computational Solutions, Lincoln University, Lincoln 7647, New Zealand Associate Editor: Martin Bishop Summary: Over past decades, constraint-based modelling has emerged as an important approach to obtain referential information about mechanisms behind biological phenotypes and identify physiological and perturbed metabolic states at genome-scale. However, application of this novel approach to systems biology in biotechnology is still hindered by the functionalities of the existing modelling software. To augment the usability of the constraint-based approach for various use scenarios, we present ORCA, a Matlab package, which extends the scope of established Constraint-Based Reconstruction and Analysis metabolic modelling and includes three unique functionalities: (i) a framework method integrating three analyses of multi-objective optimization, robustness analysis and fractional benefit analysis, (ii) metabolic pathways identification with futile loop elimination and (iii) a dynamic flux balance analysis framework incorporating kinetic constraints. Availability and implementation: ORCA is freely available to academic users and is downloadable from https://sourceforge.net/ projects/exorca/; a mini-tutorial is supplied in the package for training purposes as well as a software manual. Contact: Supplementary information: Supplementary data are available at Bioinformatics online. Received on October December 9, 2013 1 3, 2013; revised and accepted on INTRODUCTION The recent advance in high-throughput technology has vastly increased the availability of the biological information, for reconstruction of genome-scale metabolic networks (GSMs). These can serve as a detailed representation of biological reaction networks and their functional states. For investigating GSMs in both academic and industrial settings, the Constraint-Based Reconstruction and Analysis (COBRA) software (Schellenberger et al., 2011) has been established as a popular tool since 2007. Nevertheless, the systematic application of GSMs for modeldriven discovery has much scope to be further developed. For example, although the original COBRA toolbox does include a variety of methods for modelling GSMs, it still lacks ready-made functions for users to model the competition between multiobjectives, commonly encountered in bioengineering. This indicates that there is an urgent need to develop novel simulation *To whom correspondence should be addressed. 584 methodologies and tools, to interrogate and interpret the information in the GSMs in integrative, systemic and meaningful ways. To facilitate this emerging direction in research, we have developed a number of MATLAB functions, collectively packaged and named as ORCA (mOdel-dRiven disCovery and Analysis) toolbox, supplementary to the existing COBRA functions. These ORCA functions are intended to take advantage of existing COBRA models, for us to (i) obtain information on the capability of an interesting microorganism for desired metabolite production, (ii) understand the correlation between the perturbation of the surrounding metabolism, and the preferable metabolite synthesis, and (iii) investigate how dynamic changes of external metabolite concentrations in the environment affect the yield of desirable metabolites in a reactor. The tool has been used to elucidate the inherent potential of several promising biocatalysts for microbial fuel cell electricity production and some of the base mathematics of ORCA have been discussed in the literature (Mao and Verwoerd, 2013a, b, c). 2 FEATURES ORCA conducts flux balance analysis (FBA)-based methods with multi-objective formulation and extends the robustness analysis of COBRA into a framework context (Fig. 1). It incorporates a simple algorithm excluding the futile fluxes encountered during FBA and allows identification of the pathways pertaining to the metabolite of interest. ORCA also includes an updated version of the ‘dynamicFBA (dFBA)’ of COBRA. The incorporation of kinetic properties of the environment nutrient uptake was proposed in the literature (Mahadevan et al., 2002). In our updated version, limitations on biomass growth resulting from, for example, nutrient depletion can be incorporated by rateof-change constraints and substrate uptake rates modelled by Michaelis–Menten or Hill-equation kinetics. The new ‘dFBA’ function allows for unlimited number of substrates and products to be taken into account by automatically forming corresponding ordinary differential equations (ODEs). A listing of individual functions and a full discussion of how ORCA complements COBRA are provided in the Supplementary file. 2.1 Reconciliation functions Because several modifications are to be accommodated in the COBRA model before the ORCA modelling, we have included a number of reconciliation functions to help set up modelling environments, such as the formulation of multi-objectives and ß The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: ABSTRACT ORCA setting proper uptake rates of external metabolites for different cultivation conditions. In addition, ORCA includes several functions that compare models between COBRA and Optflux, another highly recognized FBA modelling software (Lakshmanan et al., 2012). This helps to resolve a different tolerance in accepting various SBML formats, as we encountered in the use of Optflux and COBRA. 2.2 Core function 1: optimizeM 2.3 Core function 2: fatmin This function iteratively implements our previously published algorithm Flux variability analysis with target flux minimization (FATMIN) (Mao and Verwoerd, 2013c). FATMIN is devised to eliminate futile loops in a metabolic network and characterize all the alternate optimal solutions or equivalent phenotypic states in a metabolism. In particular, FATMIN results characterize the network mechanisms underlying the robustness due to alternative metabolic pathways. If a series of flux levels for synthesis of desirable molecules spanning from the lowest (control state) to highest are used in running FATMIN, the function can scan for potential metabolic switches in the metabolism. These metabolic switches represent the transitions between phenotypic behaviours and can provide indirect information about a system’s tolerance to different levels of perturbation. The distinct phenotypic behaviours also indicates the suitability of the microorganism for desirable chemical production. Furthermore, the resultant lists of reactions from FATMIN can be visualized in a context of pathways and subnetworks by calling a recently published COBRA exten (...truncated)