Bofdat: Generating Biomass Objective Function for Constraint-Based Metabolic Models from Experimental Data

The computation of growth phenotypes by genome-scale metabolic models (GEMs) requires the definition of an objective function. To simulate growth related phenotypes a biomass objective function (BOF) that includes key biomass components of the cell such as the major macromolecules (DNA, RNA, proteins), lipid composition, coenzymes, inorganic ions and specie-specific components. While the accurate definition of the BOF was shown to provide precise phenotypic predictions, no standardized computational platform is available to generate species-specific data-driven BOFs. To fulfill this gap in the software ecosystem, we implemented BOFdat: a Python package for the definition of Biomass Objective Function from experimental data. BOFdat is a 3-step process. At each step, different macromolecular categories are added along with their macromolecular weight fractions, facilitating the mass balancing of the objective function. The first step adds the main macromolecules and lipids using both omics data and macromolecular weight fraction of each category to calculate the stoichiometric coefficients of each metabolite. The second step determines coenzymes by performing a connectivity analysis of the entire metabolic network, and inorganic ions by comparing against a list of ions found in all previously generated metabolic models in the BiGG Models database. The third step is a novel, unbiased approach using a genetic algorithm to find the combination of metabolites that best match experimental gene essentiality data, and clusters the results into sets of metabolic end goals. We used BOFdat to reconstruct the BOF of the Escherichia coli model iML1515 and found that the BOF generated by BOFdat had more similar biomass composition, growth rate, essentiality to the original predictions than other available methods. By providing an unbiased, data-driven approach to defining biomass objective functions, BOFdat has the potential to improve the quality of new genome-scale models and also greatly decrease the time required to generate a new reconstruction.