up regulated and down regulated genes in co expression network

Question

[izsyed16]

Hi all, I am making a co expression network using gene micro array data via Pearson correlation. If I consider a threshold value of 0.95 for correlation coefficient then my question is that should I consider positive correlation coefficient or both (positive and negative correlation values) for making network.

Answer 1

[Jean-Karim Heriche]

Consider what the correlation means. A positive one means the genes vary linearly in the same direction, i.e. they are both up or down-regulated whereas a negative one means the genes vary linearly in opposite direction, i.e. when one is up-regulated the other is down-regulated and vice versa. Now, ask yourself what is relevant to your biological question.

Answer 2

[munizmom]

Depends on which question you are interested on answering. If you are interested in identifying only the genes that are positive correlated or negative correlated or both...are you interested in identifying genes whose expression is upregulated or down regulated but follows the same direction (positive correlation) or are you interested in identifying genew that are negative correlated so in one condition are up regulated and in another downregulated?

[Arindam Ghosh]

".are you interested in identifying genes whose expression is upregulated or down regulated but follows the same direction (positive correlation)" Would it be logical construct a co-expression network of only upreg or downreg genes instead of accommodating genes of both types?

Answer 3

[izsyed16]

Can I consider both scenarios simultaneously that is can I take positive correlation (up regulated and down regulated simultaneously) and negative correlation (up regulated versus down regulated inversely) in one network that is can I take threshold as follows, correlation>0.9 and correlation <-0.9(more negative). Is it logically correct?? and what is biological relevance of such gene co expression network. If network is Un-directed?

[Jake Warner]

You can use the absolute value of the correlation. This is an 'unsigned' network. Logically it is correct but the interpretation changes as others have pointed out.
You can read more here: http://www.peterlangfelder.com/signed-or-unsigned-which-network-type-is-preferable/