SBMLKinetics: A Tool for Annotation Independent Data Driven Classification of Reaction Kinetics for SBML Models

Background: Reaction networks are widely used as mechanistic models in Systems Biology to reveal principles of biological systems. Reactions are governed by kinetic laws that describe reaction rates. Selecting the appropriate kinetic laws is difficult for many modelers. There exist tools that attempt to find the correct kinetics law based on annotations. Here, we develop annotation independent technologies that assist modelers by focusing on finding kinetic laws commonly used for similar reactions. Results: Recommending kinetic laws and other analyses of reaction networks can be viewed as a classification problem. Existing approaches to determining similar reactions rely heavily on having good annotations of chemical species, a condition that is often unsatisfied in model repositories such as BioModels. We develop an annotation independent data driven approach to find similar reactions via reaction classification, and demonstrate our approach using BioModels. Our two dimensional kinetics classification scheme (2DK) analyzes reactions along the dimensions of kinetics type (K type) and reaction type (R type). We identify approximately 10 mutually exclusive K types, including zeroth order, mass action, Michaelis-Menten, Hill kinetics and others. R types are organized by the number of reactants and the number of products in reactions. We construct a tool, SBMLKinetics, that inputs a collection of SBML models and then calculates reaction classifications as the probability of each 2DK class. We apply our tool to BioModels to compare the kinetics of signaling networks with the kinetics of metabolic networks, and illustrate significant difference of K type distributions. Conclusions: 2DK has many applications. It provides an annotation independent, data driven approach to recommending kinetic laws by using K type common for the kind of model in combination with the R type of the reaction to recommend kinetic laws. Alternatively, 2DK can also be used to alert users that a kinetic law is unusual for the K type and R type. Last, 2DK provides a way to analyze groups of models to compare their kinetic laws as we did with signaling and metabolic networks in BioModels. ### Competing Interest Statement The authors have declared no competing interest.