CARNIVAL examples
In the CARNIVAL package, 3 built-in examples are available as the test cases as follows:
- Toy Model of two crosstalk pathways
- SBVimprover species translational dataset with EGF as the perturbator
- TG-GATEs dataset with paracetamol (APAP) as the perturbator
The first example is a small toy model of 6 nodes and 6 interactions. There are two input nodes (P1 and P2) and two measured nodes (G1 and G2) with two intermediate nodes in the PKN (TF1 and TF2).
In this case study, both input nodes are set to 1 (P1=P2=1) and both measurements are also set to 1 (G1=G2=1). There are two network solutions in this case study: 1) P1 -> TF1 -> G1 and G2 (doubled activation path); 2) P2 -| TF2 -| G1 and G2 (double inhibition part). The DOT figure displays both network solutions on the same image.
The second example is one of the benchmark case using the SBVimprover dataset with EGF as the perturbator (input). We used the network from Omnipath as PKN and top 50 transcription factors (TFs) from DoRothEA together with pathway scores from PROGENy were also included as the inputs.
In this case study, the input i.e. EGF is set to 1 (as the positive regulator) linking the signalling pathway downstream towards the TFs which are changed according to the perturbation. An example of the resulting network shows that the standard ERK-MAPK signalling was not identified in this particular solution but alternative pathways towards cell cycle regulation e.g. MAX and E2Fs via CDKs as well as TGF-beta signalling (SMADs) were captured. This example highlights that there are many alternative paths that could link input(s) to measurements e.g. via crosstalks and CARNIVAL pipeline is capable of capturing such phenomenon.
The third example is an example of application study on the toxicity effects of drug (here Paracetamol/APAP) on cellular signalling process. As the toxic-related targets of drug remains unknown (in contrast to the therapeutic effect of drugs which acts on prostaglandin receptors PTGS1/2), the Inverse CARNIVAL pipeline is selected to perform the contextualisation of the network.
In this case study, the input was not provided so downstream transcription factors which are altered from the effect of drug perturbation are connected to potential nodes upstream that could regulate the signalling pathways in between. The node 'Perturbation' was introduced just to connect all possible outermost nodes in the network in order to ensure that the node penalty will also be applied to these nodes and the node 'Perturbation' should be excluded from the analysis. Note that once there are both positive and negative edges coming to a single node in the combined solution, it means that the respective interaction could either take a positive or negative path in each individual solution depended on the state of the source node. The colors of the node represent the majority of signal though.