regarding gpec : please help

[Akanksha Gupta]

i have used gpec and have certain questions:

1. while doing crossvalidation, i have specified my seed genes as the training set and all other genes in my network as the test set and achieved an auc as 0.67.

is it acceptable auc value?

2.how can i improve my auc?

3. if i give high bc and high degree nodes in my network as training set and rest others as test set, i achieve an auc of 0.98. Is this strategy of defining high BC and degree nodes as training set acceptable?(as high BC and high degree nodes are bottlenecks and hubs of network and are bound to be involved in the disease, i guess...)

[Duc-Hau, Le]

Hi,

I am the author of GPEC. Thanks for using our plugin.

For your questions:

1. AUC = 0.67 can be considered acceptable value, since the random value will be 0.5. The AUC value generally depends on how much close to each other in the network genes associated to a disease are. 

2. Therefore, if you want to improve the AUC value, you should use more dense network where genes associated to the same disease are expected to be close to each others. For this, a functional linkage network should be used.

3. You are right. Actually, RWR algorithm (the base algorithm of GPEC) is biased to hub genes (high degree genes). Fortunately, disease genes are naturally located in hubs in interaction networks and genes associated to the same disease usually form a so-called disease module.

Hope this helps,

Cheers,

--
You received this message because you are subscribed to the Google Groups "cytoscape-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cytoscape-disc...@googlegroups.com.
To post to this group, send email to cytoscap...@googlegroups.com.
Visit this group at http://groups.google.com/group/cytoscape-discuss.
For more options, visit https://groups.google.com/groups/opt_out.

--
Duc-Hau Le, Ph.D (Vietnamese: Lê Đức Hậu)
Cell phone: (84) 912-324564 (Vietnam), (82)10-6869-2999 (Republic of Korea)
Email:haul...@gmail.com, haul...@wru.vn
Website: https://sites.google.com/site/duchaule2011/
Faculty of Computer Science and Engineering, Water Resources University.
175 Tay Son, Dong Da, Hanoi, Vietnam.

[Akanksha Gupta]

Hello sir,

Thankyou for your previous reply. Your reply helped me in getting my gpec analysis work published in a reputed journal. Now, I wish to perform some more similar kind of analysis using gpec and wish to generate roc curve for the same. But, it says 'cannot find the path specified' and hence does not calculate auc or generate roc curve. Now, when I tried uninstalling and reinstalling the cytoscape 2.8.3 version again on my workstation, I am unable to install the gpec app. Please help since it a very urgent.

Thankyou in Advance