Question

Cutoff for coexpressed gene pairs using Pearson/Spearman correlation

0

Entering edit mode

juls • 0

@juls-11275

Last seen 3.4 years ago

Austria

Dear all,

I am using the cor.test function (Spearman/Pearson correlation) to find coexpressed genes for a specific gene of interest in my microarray dataset (226 samples).

My question is now - what is the established cutoff for coexpression? Just the p-value results in a large number of coexpressed genes. I also categorised the genes according to correlation coefficient as in this list (see below), but is there any established cutoff > 0.5 or is indeed mainly the p-value used?

0-0.19	very weak
0.20-0.39	weak
0.40-0.59	moderate
0.60-0.79	strong
0.80-1.0	very strong

Thanks for advice!

Julia

microarray pearsoncorrelation coexpression spearman • 2.0k views

ADD COMMENT • link updated 7.8 years ago by James W. MacDonald 67k • written 7.8 years ago by juls • 0

score 0 · Answer 1 · 2017-03-03

0

Entering edit mode

James W. MacDonald 67k

@james-w-macdonald-5106

Last seen 1 day ago

United States

What about negative correlation? Is that not co-expression of a type as well?

Anyway, I am not sure inference will help you here, as a p-value doesn't measure the strength of correlation. Instead it just says how likely your result would be if the two genes really were not correlated. With 226 samples, even a pretty small correlation will have a p < 0.05 (somewhere around 0.11 or so). You will probably just have to set some ad hoc cutoff that you define as 'correlated enough' and go with that.

ADD COMMENT • link 7.8 years ago James W. MacDonald 67k

0

Entering edit mode

Thanks for your answer.

I adjusted the p value, so it's around 0.18 but yes it's very low. Hence my question about an "established" cutoff. Any suggestions about a sensible cutoff here?
(The above list was also meant for negative correlation of course. I classified with the abs(correlation). I thought that was clear ;) - sorry)

ADD REPLY • link 7.8 years ago juls • 0