which algorithm to use for clusterviz analysis

[ekta]

I downloaded Clusterviz. For the same network; I run all FAG-EC,
MCODE and EAGLE options and also Nemo. As you can guess, they all give
different modules. Modularity (i guess) is the important parameter
here. Or should i take all results into consideration regardless of
their size/modularity/score?

[Scooter Morris]

Hi Ekta,
I can't give you much advice about Clusterviz, but I can give you a
number of suggestions regarding clustering in general, and clusterMaker
in particular. The "right" answer is that "it depends". You can tune
your clustering based on a modularity score or some other metric, but
that only makes sense if that score correctly reflects your biological
insight into the underlying processes. You also don't say whether your
network is weighted or unweighted. Assuming you have a weighted,
undirected network, I've had good success with MCL. The latest
clusterMaker also supports a number of post-clustering filter options
(like MCODE does) that might be helpful. If you are going to compare
methods, the best way is to use a set of "knowns", but otherwise, I
would certainly choose a metric (e.g. Modularity) and choose the
algorithm that gives you the best modularity score.

-- scooter

[Ekta Manglesh]

thanks Morris. MCL has given me expected results :)

--
smile nd d wrld smiles :)))