Diffany: an ontology-driven framework to infer, visualise and analyse differential molecular networks

Sofie Van Landeghem1,2 , Thomas Van Parys1,2, Marieke Dubois1,2, Dirk Inzé1,2 and Yves Van de Peer1,2,3*

1 Department of Plant Systems Biology, VIB, Technologiepark 927, 9052 Gent, Belgium.

2 Department of Plant Biotechnology and Bioinformatics, Ghent University, Technologiepark 927, 9052 Gent, Belgium.

3 Genomics Research Institute (GRI), University of Pretoria, Private bag X20, Pretoria, 0028, South Africa.

*Corresponding author, E-mail: yves.vandepeer@psb.vib-ugent.be

Abstract

Motivation:
Differential networks have recently been introduced as a powerful way to study the dynamic rewiring capabilities of an interactome in response to changing environmental conditions or stimuli. Currently, such differential networks are generated and visualised using ad hoc methods, and are often limited to the analysis of only one condition-specific response or one interaction type at a time.

Results:
In this work we present an ontology-driven framework to infer and visualise an arbitrary set of condition-specific responses against a reference network. To this end, we have implemented algorithms that can process highly heterogeneous networks, accounting for both physical interactions and regulatory associations, symmetric and directed edges, edge weights and negation. We propose this integrative framework as a standardized methodology that allows a unified view on differential networks and promotes comparability between differential network studies. As an illustrative application, we demonstrate its usefulness on a plant abiotic stress study and experimentally confirmed a predicted regulator.

Availability:
Diffany is freely available as open-source java library and Cytoscape plugin from http://bioinformatics.psb.ugent.be/supplementary_data/solan/diffany/.

Reference:
Van Landeghem, S., Van Parys, T., Dubois, M., Inzé, D., Van de Peer, Y. (2016) Diffany: an ontology-driven framework to infer, visualise and analyse differential molecular networks. BMC Bioinformatics 17:18

Figures

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Figure 1 presents an artificial example illustrating the application of Diffany to infer differential (c) and consensus (d) networks from heterogeneous reference (a) and condition-specific (b) networks. Edge weights are encoded in the edge thickness and also displayed next to the interaction type. Notice how directionality, different edge types and weights can all be mixed freely in the networks, illustrating the power of our ontology-based framework when applied to such heterogeneous input networks.

The resulting differential network summarizes the rewiring events that occur when a certain environmental condition or stimulus is introduced, while the consensus network models those interactions that remain unperturbed.

The application of our framework to much larger networks, reanalysing a previously published experimental dataset on abiotic stress reponses in Arabidopsis, is detailed in our manuscript. In this study, we have constructed four condition-specific networks modeling a mannitol-induced stress response at 1.5, 3, 12 and 24 h after transfer of 9-days old seedling to a mannitol-containing medium. With Diffany, we can visualise the stress response at a specific time-point (e.g. at 1.5h, cf. Figure 2), focusing only at those specific rewiring events. Further, we can compare the reference network to all four time points simultaneously, and calculate the overall differential rewiring common to all time points (Figure 3). Finally, a more robust network inference is obtained by applying a less stringent criterium which only requires that three out of four time points share the same rewiring pattern for it to be included in the differential network (Figure 4, showing only regulatory associations).

Legend figures 2,3 and 4:
The edge color denotes the rewiring event: increase/decrease of regulation in dark green and red respectively, increase/decrease of PPI in light green and orange, increase/decrease in phosphorylation in blue and purple. It is important to note that in these differential networks the arrows point to rewiring events: a decrease of regulation for instance (red arrows) does not necessarily point to an inhibition, but may also indicate a discontinued activation. Diamond nodes represent proteins with a known phosphorylation site, and proteins with a kinase function are shown with a black border. Blue and yellow nodes identify underexpressed and overexpressed genes respectively.

Downloads


Contact:
VIB / UGent
Bioinformatics & Evolutionary Genomics
Technologiepark 927
B-9052 Gent
BELGIUM
+32 (0) 9 33 13807 (phone)
+32 (0) 9 33 13809 (fax)

Don't hesitate to contact the in case of problems with the website!