[back to the DMCM’24 Programme]

Title:
Analysis of Partially-Specified Boolean Models with BioDivine AEON Framework

Authors:
David Šafránek, Samuel Pastva, Lubos Brim, Nikola Benes and Ondřej Huvar

Abstract:
An important issue arising in Boolean Networks (BNs) is that their behaviour is often subject to the influence of partially or entirely unknown update functions. For example, the lack of information on update functions arises when transforming SBGN schemes to Boolean Network models. The partially specified update functions can be technically handled using the so-called uninterpreted Boolean functions. The uninterpreted functions can be understood as parameters of the partially-specified models.

In our computational framework, we address two problems targeting partially specified models: the first one is the attractor identification, i.e. the computation of states towards which the network eventually converges, together with the corresponding parameter settings which enable these attractor states. The second problem is source-target control, which identifies network perturbations and parameter settings that guarantee that the desired (target) attractor is achieved.

In this presentation, we demonstrate the functionality of our software framework BioDivine AEON that comprehensively addresses these problems for partially specified BNs with asynchronous update. The framework can handle reasonably large BNs due to its symbolic algorithms based on binary decision diagrams (BDD) representation of the system’s dynamics. To visualise the results of the attractor analysis, an interactive presentation technique based on decision trees is employed. It allows users to quickly explore parameters that can cause crucial changes in long-term dynamics.

The variable stability analysis feature provided in our tool allows to use automatically generated decision trees to explore the system’s dynamics for presence/absence of particular biological phenotypes. This helps the user to reveal the effect of input variables (e.g., environment stimuli or drugs) on the studied biological phenotypes.

The key features of the framework will be demonstrated on an abstract model obtained from COVID19 Disease Maps project. In particular, the presented example will represent a complex human cell signalling pathway describing the activity of type-1 interferons and related molecules interacting with SARS-COV-2 virion. The analysis focuses on explaining the potential suppressive role of the known drug molecule GRL0617 on replication of the virus.

Our presentation can be realised both as a talk or as a poster. The presented information will be based on the following papers:

Benes et al. (2022) Bioinformatics 38 (21), 4978-4980 Benes et al. (2022) BMC Bioinformatics 23 (1), 173