Error in exporting igraph modules

Question

[anikng]

I was trying to identify modules from RNASeq count data with help of igraph tutorial. It didnt throw any error in Graph and tree object generation, and community identification with cluster_louvain() method. I almost used the same code shown in the tutorial.However, when I tried to output the community information, it throws error

commSummary <- data.frame(
 mst.communities$names,
 mst.communities$membership,
 mst.communities$modularity)
 colnames(commSummary) <- c("Gene", "Community", "Modularity")

Error in data.frame(mst.communities$names, mst.communities$membership, : arguments imply differing number of rows: 17965, 3

From stackoverflow solution for a similar problem, it looks like n_rows!= n_cols is the problem. But it is not the case here. Can anyone suggest a solution for this error?

dim(df))

[1]  17966     6

head(df)

               8hc    9hc    12hc     8hs     9hs   12hs
02m156150   -4.275  9.528   -3.061  -3.17   0.134   9.659
02m256640   -4.025  10.005  -2.21   -2.656  1.142   9.397
02m146600   -3.154  8.553   0.444   -3.728  1.394   10.909
03m313660   -2.139  10.54   -1.644  -2.103  0.923   8.965
02m151678   -4.025  8.164   -2.323  -3.768  -1.609  7.888

If required, I can paste the whole code that I used.

Answer 1

[Kevin Blighe]

Hi, I wrote that tutorial. You could simply check the contents of these vectors to see what went wrong.

mst.communities$names
mst.communities$membership
mst.communities$modularity

They should have equal length, but it can happen in rare occasions where they don't. The likely cause is that one of your vertices (genes) has no edges (no connectons) - this can occur where all of its edges fell below the threshold that I use for filtering in the tutorial.

Kevin

[anikng]

Hi Kevin, thanks for the reply. I checked output of those 3. Length of mst.communities$names and mst.communities$membership are same.However I noticed that most of the genes in the mst.communities$membership is assigned to 1st module and remaining to 2nd and 3rd module. Output of mst.communities$modularity, I dont know if that is a length problem.

    x
 1  0.032290429
 2  0.032734522
 3  0.032836322

I tried the same code with a subset of 2000 genes (ie, only TF genes) an it worked well. Do you have any suggestion in this case?

[Kevin Blighe]

If they are of the same length and are each a vector, then there should be no issue with that command. Can you confirm the output of:

str(mst.communities$names)
str(mst.communities$membership)
str(mst.communities$modularity)

?

Regarding the finding that the majority of genes are in the same module, that could reflect the structure of your data - is it non-uniform (heteroskedastic)? If your data is RNA-seq, the input should have been transformed, normalised data.

[anikng]

Yes, I used normalized data from RNASeq. Tried with both normalized count data as well as log2 format. But both gave same result. I checked the output and it looks like,

str(mst.communities$names) chr [1:17966] "02m151500" "02m256640" "02m141600" ... str(mst.communities$membership) num [1:17966] 38 43 100 45 30 30 55 64 80 134 ... str(mst.communities$modularity) num [1:6] 0.615 0.843 0.938 0.974 0.984 ...

[Kevin Blighe]

Oh, the modularity vector only has a length of 6. That may be some change introduced in igraph, but not sure...

You can definitely combine names and membership into a data-frame, though (they have equal length).

data.frame(
  names = mst.communities$names,
  membership = mst.communities$membership)

Another point to note is that you seem to have a vast number of modules / communities... it can be difficult to work with that amount.