I am working with a dataset that includes only 2 conditions. Essentially, a treatment and a control condition. I would like to include genes that are positively and negatively correlated with one another into the same modules ("unsigned") in order to preserve potential inter-gene regulations for downstream functional analyses. However, given that this is a binary dataset and I am using an unsigned network, I don't know how to correctly interpret the module correlations with the treatment groups (i.e., "trait" data).
As you might expect, if a module has a large, significant correlation with one condition of, let's say, cor. +0.75, then the direction of this correlation will be exactly opposite in the other treatment group, so cor. -0.75. But from my understanding, it would not be correct to say that the positively correlated module has a greater association with that particular condition because it contains unsigned genes. Therefore, the only conclusion one could make would be that the correlation itself is significant (simply, the module is impacted in some way by the treatment), but say nothing about directionality or prevalence of the genes contained within that module. In other words, you can't really say that the module is down or upregulated in the treatment group vs. the control, just that the module is differentially regulated.
Does this interpretation make sense? Or am I totally off the mark on this??
It would be great to get some expert input here. Thank you! -Peter