I have a network with 500 nodes and 400 edges. Iam interested in the pathways between 2 nodes and for that i have calculated all the shortest paths between that 2 nodes. To test experimentally, i need to select/predict "important" pathways, as testing all shortest paths will be laborious, and right now Iam using graph-theoretic measures like betweenness centrality, maximum indegree and outdegree of the nodes in my shortest paths and filtering the ones which have lower value.
Do you think this is enough/appropriate or are there any other ways that i can select/predict the "important" pathways which i can test experimentally?
My firend, Esti Yeger-Lotam developed this:
http://netbio.bgu.ac.il/respnetnew/
What do you mean by "important"? And in what context? In disease, the absence of a biological pathway link can be more important than its presence; this is found by comparing the pathway network(s) to the normal or control pathway network(s). More network analysis tools can be found in computer science and operations research (engineering) if you're looking for something that is "important" for other reasons.
Just an idea: in a natural network like metabolic pathways
vital paths are often redundant. If a randomly chosen edge
is removed on the route from node n1 to node n2, there
will with high probability still be a path from n1 to n2,
and its length will roughly be equal to the former one.
This effect could be used a basis for measuring path
importance. The idea is to repeatedly modify the graph
by removing edges, eventually starting from the
original graph, and computing the average length
of the path between n1 and n2.
Regards,
Joachim
Thank you for the replies and suggestions.
I would like to explain in mode details the scope and context:
The context is not disease but the stimulation of cells with a chemical that triggers TLR responses:
The network which i have is derived from literature using the tool ARIADNE (Pathway Studio) and also the predictions of transcription factors (from TRANSFAC) for my genes of interest. As i am interested in finding the potential pathways between 2 nodes, i have calculated all paths between those nodes in my network using "graphkshortestpath" toolbox from MATLAB. It gave me roughly around 2000 paths that can be between the 2 nodes.
One way that i tried was similar to Mr. Joachim Pimiskern suggestion of calculating the essentiality of each gene in contributing to the pathway between the 2 nodes. This is the way i had selected some genes and will do a perturbation experiment to see if the genes effects the paths. The problem here was the gene which im going to test are involved in lets say 30 paths each between the 2 nodes. I can now continue the same procedure as explained before for after testing every gene but this will too long and its like a classical procedure (i feel). Hence i would like to know if there are some other ways to do this.
Thank you
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Bioinformatics
- Bioinformatic Tools