Repairing Boolean regulatory networks using Answer Set Programming

Abstract

Models of biological regulatory networks are increasingly used to formally describe and understand complex biological processes. Such models are often repaired whenever new observations become available, because the model cannot generate behaviours consistent with the new observations, or because the behaviours are contradictory. This process of model repair is often manual and therefore prone to errors. In this work, we describe biological regulatory networks using the Boolean logical formalism, where nodes are represented by Boolean variables denoting biological components and edges denote regulatory interactions between components. The evolution of each variable is defined by a Boolean logical function depending on the values of the regulators of the component. Here, we propose to repair the model by changing inconsistent functions, with four types of atomic repairs which can be further combined. The goal is to find the cardinality minimal set of repairs allowing the model to satisfy all available observations. The proposed method is implemented using Answer Set Programming (ASP) and is tested using data from Escherichia coli and Candida albicans organisms. Interestingly, the system finds adequate solutions to ensure consistency for all observations.

Downloads

• Paper
• Program
• Instances: Escherichia coli
• ASP: clasp + gringo

Example of usage

To use the program you should pass it an influence graph and experimental profiles, for example:

It is possible to run the program allowing the use of different repair operations:

• Repair e - removes edges
• Repair i - negates the regulators of a function
• Repair n - negates a function
• Repair g - changes AND to OR and NOT to the identity function.

Note: All the combinations of repairs are allowed, and they need to be written in alphabetical order.

Contacts

If you have any comments or questions, please .