Hi,
For constructing network I used WGCNA alghorithm.So,before constructing network, I have to determine SFT index. for that, I ran below code:
sft = pickSoftThreshold(MyData, powerVector = powers, verbose = 5) So I get 10 option as below: Power SFT.R.sq slope truncated.R.sq mean.k. median.k. max.k. 1 1 0.439 -7.21 0.715 3280.000 3210.000 4890.0 2 2 0.931 -5.79 0.970 614.000 579.000 1460.0 3 3 0.974 -4.44 0.991 150.000 133.000 602.0 4 4 0.968 -3.61 0.984 44.600 36.600 304.0 5 5 0.957 -3.08 0.978 15.700 11.600 176.0 6 6 0.943 -2.72 0.972 6.420 4.140 112.0 7 7 0.922 -2.48 0.960 2.980 1.620 76.2 8 8 0.926 -2.25 0.964 1.550 0.688 54.4 9 9 0.922 -2.09 0.952 0.887 0.310 40.4 10 10 0.958 -1.88 0.957 0.552 0.147 30.9
so, after that for determining power value I have to plot based on below code:
par(mfrow=c(1,2)) # SFT index as a function of different powers plot(sft$fitIndices[,1],sign(sft$fitIndices[,3])*sft$fitIndices[,2], xlab="Soft Threshold (power)",ylab="SFT, signed R^2",type="n",main=paste("Scale independence")) text(sft$fitIndices[,1],sign(sft$fitIndices[,3])*sft$fitIndices[,2],labels=powers,col="red") # this line corresponds to using an R^2 cut-off of h abline(h=0.90,col="red")
My problem is in Scale independence plot. all power values are upper 'abline' and my abline has no intersect to any value. Now, based on this situation for constructing network, I have to consider maximum 'SFT.R.sq' as selected power or not? if this is true, should I select my power equal 3?
I appreciate if anybody share his/her comment with me.
Best Regards,
Mohammad
P.S: my network is unsigned.
That's an interesting problem! Finding the optimal soft power threshold in WGCNA can definitely be tricky. I've found myself experimenting quite a bit to find a good balance between scale-free topology and maintaining sufficient connectivity. It's almost like playing a little optimization game. Sometimes I feel like I'm dodging cacti in the Dinosaur Game, constantly adjusting parameters to avoid a premature crash! I wonder if anyone has explored using more automated or adaptive methods for this selection process to improve reproducibility?
Exploring the challenges of selecting the right soft power in WGCNA highlights the delicate balance between sensitivity and specificity when building gene co-expression networks. I once faced a similar issue during my research on plant stress responses, where choosing the soft threshold significantly influenced the network topology. Tools like those from Monkey Mart can sometimes offer innovative solutions or guidelines to simplify this complex process.