SBMLlint

Problem Addressed

Many biological models are based on chemical reactions. For example, glycolysis, arguably the most widely exercised metabolic pathway in biology, begins by transforming the reactants glucosue (Glu) and adenosine triphosphate (ATP) into the products glucose 6-phosphate (GluP) and adenosine diphosphate (ADP), or: Glu + ATP -> GluP + ADP. Examples of biological modeling techniques that rely on reactions include flux balance analysis and kinetics models.

Today's biological models typically consist of tens to hundreds of reactions. Even with this modest level of complexity, it is easy to make mistakes. For example, a mass balance error occurs if the total mass of the reactants differs from the total mass of the products. With the advent of high throughput laboratory techniques, the complexity of models will grow rapidly. As a point of comparison, the complexity of typical software packages has grown from hundereds of lines of code in the 1960s to tens of millions of lines of code for software such as linux and the Apache web server.

Because of this huge growth in the complexity, software engineers developed sophisticated tools to detect errors in codes statically, before any statement is executed. For example, the pylint tool analyzes python source codes to determine if a variable is referenced before a value is assigned to it. The term linter is used for a tool that does static analysis of source codes.

Overview

SBMLLint is a collection of tools for linting reactions. The initial focus is detecting mass balance errors. The tool takes as input a model expressed in either SBML (Systems Biology Markup Language, a standard format for biochemical models) or the Antimony language (a human readable representation of chemical reaction models).

SBMLLint implements two algorithms for detecting mass balance errors. The first, moiety analysis, checks for balance in the moiety structure of reactions. For example ATP has the moeities adenosine with three inorganic phosphates. Moiety analysis requires that modelers follow a naming convention that exposes the moiety structure. There is no restriction on the choice of moiety names (other than compliance with SBML naming standards). Moiety names can be exposed in two ways. The first is by using a naming convention. For example, the modeler could use A to indicate a adenosine moiety and Pi for inorganic phosphate. Moiety analysis requires that moieties be separated by an underscore (\_). That is, ATP would be written as A_Pi_Pi_Pi Similarly, GluP would be written as Glu_Pi. Thus, the above reaction is written as Glu + A_Pi_Pi_Pi -> Glu_Pi + A_Pi_Pi. A second way to expose moiety names is through explicit declarations in a configuration file. An example of this is

ATP:
- A, 1
- Pi, 3

We provide a tool to partially automate the construction of these explicit declarations (make_moiety_structure).

Moiety analysis checks that the count of each moiety in the reactants is the same as the count of each moiety in the products. Although moiety analysis places a burden on the modeler to use the underscore convention, we note that about 20% of the models in the BioModels repository already use names that are close to this structured.

The second algorithm, GAMES (Graphical Analysis with Mass Equality Sets) does not impose any requirements on the structure of the molecule names. GAMES checks for stoichiometric inconsistency, which is a weaker condition than mass balance. A collection of reactions is stoichiometrically inconsistent if the set of reactions infers that a molecule has a mass of zero. To illustrate this, consider a model consisting of two reactions: A -> B and A -> B + C. The first reaction implies that the mass of A is the same as B. The second reaction implies that the mass of A equals the sum of the masses and B and C. Both statements can be true only if the mass of C is zero, and so the model has a stoichiometric inconsistency.

Using the Tools

SBMLLint provides the following command line tools that are available when SBMLLint is installed.

• moiety_analysis uses moiety analysis on an SBML source file to report on mass balance errors.
• games uses games on an SBML source file to report on mass balance errors.
• make_moiety_structure takes as input an SBML XML file and a YAML file that lists moiety names to construct the YAML for explicit declarations of moieties.
• print_reactions takes as input an SBML XML file and prints the reactions in the model (including their kinetics)

The following is an example of using the moiety_analysis and ``GAMES` algorithms to check for mass balance in a Jupyter Notebook.

SBMLLint can also be run from the command line, taking as input a model file expressed in SBML or Antimony (if you install Tellurium). Below are examples (although the outputs have been truncated).

For moiety analysis:

For GAMES:

Below we illustrate the use of make_moiety_structure and the format of the YAML input file and the output produced.

Here is a an example of using print_reactions.

Installation

SBMLLint is distributed through PyPI. You can install using pip install SBMLLint.

To verify the installation:

1. Clone the repository using git clone https://github.com/ModelEngineering/SBMLLint.git
2. nosetests SBMLLint/tests on Mac and Linux; nosetests SBMLLint\tests on Windows.

Depending on your environment, you may see some warning messages, but there should be no errors.

The pip install does not include tellurium, and so by default you cannot analyze Antimony files. If you want to analyze Antimony files, you can either install tellurium separately, or python setup_tellurium.py install in the repository.

SBMLLint configuration file

SBMLLint can optionally be used with a configuration file. An example of the file can be found in the SBMLLint github folder SBMLLint/.sbmllint_cfg.yml. This is shown below:

The sections of the configuration file are the top level tags that end in a colon (":"). The first two sections provide examples of ignoring molecules and moieties in moiety analysis. The next section indicates whether boundary reactions are considered. By default, boundary reactions are not considered since, by definition, they create or destroy mass. The last section provides for explicit declarations of moiety structures.

The configuration file can be specified in the moiety_analysis and games tools by specifying the ```--config`` option.

Development Notes

The repository contains several shell scripts that add in code development.

• install.sh installs the code in the test-sbmllint virual environment.
• setup\_env.sh sets up the environment variables (use source setup_env.sh).