[BioC] SPIA: inferring inhibited pathways from up-regulated genes?

[Tarca, Adi]

 Dear Paul,
The primary information that the spia function returns is a global impact p-value. The additional label "activated" or "inhibited" for each pathway is not decided by counting how many genes are up and how many are down in the pathway. You can easily do this if you want to take that road.
To give you a sense of how the activated or inhibited lables are assigned, assume a simple scenario in which all you have is one (N=1) differentially expressed gene which is up regulated, but this gene has inhibititory connections to many other downstream genes or activation connections to genes which at their turn are inhibiting many other genes. This will result into a net negative perturbation accumulation. However, the reference we use here is not considered to be 0, therefore some negative accumulation does not directly translate into an "inhibited" pathway. The reference we use is the expected (mean) perturbation accumulation under the null distribuition which is generated by from random scenarios where the same number of de gene IDs (here N=1) are chosen at random from the pathway and assigned random log fold changes (from all DE genes on the array).   
In terms of pointing out which genes contribute the most to the observed perturbation accumulation is not a feature that we have implemented so far. Note though that the relevance of a given DE gene depends on which other genes are DE for a given pathway.
Regards,

Adi Tarca


-----Original Message-----
From: Paul Shannon [mailto:pshannon at systemsbiology.org] 
Sent: Saturday, December 19, 2009 6:54 PM
To: Tarca, Adi
Subject: SPIA: inferring inhibited pathways from up-regulated genes?

Hi Adi,

I ran SPIA on some expression data, and SPIA tells me (among other things) that the TLR signaling pathway is inhibited.  I attach a network view below, in which red genes are up-regulated, green are down.  I only submitted the twelve upregulated (red) genes.  
color coding --- 
   log2=3: dark red.  log2=0: white.  all other values are interpolated
I'm afraid I don't understand how inhibition at the pathway level is inferred here.  Could you help me to understand that, please?  
Also: is it possible to query and/or examine the SPIA results to learn how this assessment was made?  So that we can then say something like
   This pathway is inhibited due to high expression values for these focal genes:  x, y, z.
I am misinterpreting all this, I'd be grateful if you could set me straight.
Thanks!
 - Paul Shannon