WGCNA problem

Question

[mortezaali]

Hello. I have problem to choose soft power for WGCNA. please guide me!! I have 501 sample (cancer=481 , normal=39) that download from TCGA database. Then, normalization and filtration was carried out using TCGA package. With this data I started using WGCNA for co-expression network analysis. For selecting soft threshold I see my R^2.sq is lower than 0.8 . please help me to chose best power. Thank you.

data_tcga_501_ordered_t <- t(data_tcga_501_ordered)
gsg <- goodSamplesGenes(data_tcga_501_ordered_t)

power <- c(c(1:10) , seq(from = 12 , to = 20 , by = 2))

sft <- pickSoftThreshold(data_tcga_501_ordered_t , powerVector = power ,  verbose = 5)

Power SFT.R.sq  slope truncated.R.sq mean.k. median.k. max.k.
1      1   0.9520  1.500          0.955  7150.0   7630.00  10300
2      2   0.0707  0.163          0.125  3840.0   4030.00   7220
3      3   0.1880 -0.392          0.324  2300.0   2270.00   5340
4      4   0.3680 -0.717          0.522  1470.0   1340.00   4100
5      5   0.4630 -0.929          0.630   992.0    818.00   3230
6      6   0.5300 -1.080          0.704   694.0    513.00   2600
7      7   0.5710 -1.210          0.752   500.0    329.00   2130
8      8   0.5930 -1.330          0.785   368.0    216.00   1780
9      9   0.6280 -1.390          0.821   277.0    145.00   1490
10    10   0.6530 -1.440          0.846   212.0     99.00   1260
11    12   0.7050 -1.500          0.892   129.0     48.50    922
12    14   0.7410 -1.530          0.920    81.9     24.90    688
13    16   0.7630 -1.570          0.936    53.8     13.30    525
14    18   0.7650 -1.640          0.946    36.4      7.34    411
15    20   0.7760 -1.680          0.956    25.1      4.15    326

Answer 1

[andres.firrincieli]

A table in the WGCNA faq tells you how to select the appropriate soft-thresholding power in case you can't get a good scale-free topology index: link.

However, before jumping on WGCNA you should run a PCA on the normalized expression data to check if you have outliers.

[mortezaali]

Thank you for your attention Andres I drew the PCA diagram before WGCNAenter image description here and I will upload the result along with the box plot and I would be grateful if you look at these and tell me your opinion.

[andres.firrincieli]

You have complete separation between the groups which is fine (the second plot is not really useful).

You should probably run WGCNA with the settings for a signed network:

pickSoftThreshold(data_tcga_501_ordered_t , powerVector = power ,  verbose = 5, networkType = "signed").

If you still can't get the appropriate soft-thresholding power, use power=12 instead.

Finally, remember to set type = "signed" when using the adjacency() function, and TOMType = "signed" for TOMsimilarity().

Have fun.

[mortezaali]

Thank you very much👊