Home > Community > How to select a power for a scale-free topology network
Upvote

22

Downvote
+ Bioinformatics
+ Biochemistry
+ Statistics
Posted by
Mani Bhullar

How to select a power for a scale-free topology network

Danielle Fraser  Follow

I would be concerned about the input data that produced these numbers. This dataset has a really odd shape with a dip at the start:

data.df <- read.table("soft_threshold.txt", header=TRUE);png("scale-free-194.png");plot(data.df$SFT.R.sq);dummy <- dev.off();

scale-free power with an odd inflection

Here's what I expect this to look like (from the supplementary information in this paper):

normal scale-free topology

It's easier to select the power/threshold when the plot looks like the second situation. With the data as presented, it doesn't look like any value would be suitable. but if I were forced to pick, then I'd choose 12 because it is the first number above the magic threshold (ignoring the weirdness at the start). However, I'll admit that I don't understand the concepts behind the power, and am just following the recommended instructions, so can't explain why it is what it is.

More

Upvote

VOTE

Downvote
Duane C. Woodall  Follow
Yes, I am aware of this particularity, I think it is due to a bug in the WGCNA package, when pickSoftThreshold doesn't use expression data. I probably should have picked a more typical data. But simply cutting at certain threshold leaves outside the consideration about the mean connectivity part. (Also it seems to treat the same that the slope of the power =1 is positive and the others become negative)More
Upvote

VOTE

Downvote