Inference in Biological Regulatory Networks

Abstract

Nowadays, most biological models are still handmade requiring a great amount of effort by the modeller. Different models can be derived from the same set of data and different modellers will therefore most likely build different models. Every time new data is obtained, it is necessary to reassess the model’s consistency. If the model is not consistent with the new data, then it needs to be corrected. With the large amount of data available, repairing a model by hand is often a difficult task and so it is important to reduce the difficulty of this task by creating computational tools that allow the representation of models and to reason over them. There are already some tools that can repair models. However, these tools do not frequently allow the use of Boolean functions to explain the behaviour of the biological components. In this work, different approaches to model and repair biological networks are analysed, proposing a new set of repair operations. This set was implemented using two different approaches, one using a Maximum Satisfiability (MaxSAT) solver and another using an Answer Set Programming (ASP) solver. Both tools can repair biological networks, using a logical formalism, with multiple sets of experimental data. The two tools were tested and compared using biological networks of Escherichia coli and Candida albicans. The set of repair operations proposed was able to solve all inconsistent models, and the solution implemented using Maximum Satisfiability (MaxSAT) was the most efficient.