Closed diffany-admin closed 9 years ago
Comment by svlandeg Thu Mar 6 11:04:34 2014
Now working on this with Marieke's osmotic stress data
Comment by svlandeg Fri Jun 20 09:32:43 2014
Example, original network, with DE genes presented by colored nodes:
Differential network:
Comment by svlandeg Thu Aug 28 14:29:54 2014
Discussed this with Marieke and Nathalie. The differential network should also include those interactions where only one node is DE (e.g. HY5). Perhaps with the exception of too generic transcription factors, which are sometimes overrepresented in the data because of corpus bias (e.g. AGL 15). For such "hubs" in the data, we can check the proportion of interaction partners that are in the seed set (in our case: DE genes), to determine whether or not the hubs are relevant.
Further, some sort of "fuzzyness" interval would be great for the non-DE genes which only barely didn't make the cut, e.g. between 0.05 and 0.1 -> include them if they have a high FC, or if they are connected to the existing network, etc.
Finally, we agreed on removing unspecified regulatory associations, to avoid pulling in generic TFs with only computational evidence for instance.
Comment by svlandeg Fri Aug 29 14:35:41 2014
Done: filtering unspecified regulatory interactions, thus removing much data from AGL15, GL1/3, SEP3 and HY5. The PPI data probably also contains such hub biases! (cf #192)
Issue by svlandeg Thu Oct 31 16:52:32 2013
Calculate differential network from microarray data: https://github.ugent.be/BEG/diffnet/wiki/Differential-algorithms#from-2-microarray-datasets