Improving the FAIRness of causal interactions in systems biology

Abstract

In molecular systems biology, regulatory process models are built and computationally simulated to represent and analyse the behaviour of cellular systems and how they respond to perturbations. A considerable amount of knowledge in public databases is available in the form of biological networks depicting metabolic reactions, cellular signalling cascades, or gene regulatory events that are initially assembled from molecular interactions extracted from the literature. Still, it remains challenging to convert this accumulated knowledge to mathematical frameworks that correctly capture the mechanistic details of networks. To make better use of the available knowledge, these networks can be deconstructed into their most basic regulatory network motifs, called a causal statement. A causal statement is a directed interaction where a source entity (regulator) influences the activity of a target entity (regulated entity), annotated together with contextual information. These statements describe how information flows in a signal transduction network. When provided with sufficient contextual details, they can enhance the understanding of the enabled biological mechanisms, and increase the subsequent analyses that can be performed with them. Recent scientific efforts started to capture these interactions and the collected data is now increasingly used in mathematical modelling projects. However, no common infrastructure has been put in place to manage causalities in a “FAIR” (Findable, Accessible, Interoperable, and Reusable) manner.

The thesis work aimed to answer two main questions:

how to use available prior knowledge from pathway databases to extract by inference causal statements? how to tackle the existing variability in the current representations of causal interactions, at the annotation and data management levels?

These objectives have been achieved with the establishment of curation guidelines, the implementation of a curation platform to support these guidelines, the increase of interoperability between resources and data formats by following the guidelines, and finally the inference of causality from prior-knowledge networks to make existing content reusable for modelling applications.