Dear All
I'm ytanaka.
Although we built the RNA-seq pipeline with CummeRbund, we met an error
frequently when we run the csCluster method in CummeRbund.
Please let us know how to fix this issue.
The version of the software we used is as follows.
====================
- Tophat: 2.0.13
- Cufflinks: 2.2.1
- CummeRbund: 2.8.2
- R: 3.1.1
====================
The following is the script to run CummeRbund.
=================================
$ /tool/R-3.1.1/bin/Rscript exec_clustering.R
### exec_clustering.R ###
library(cummeRbund)
cuff <- readCufflinks()
cuff_gene_ids<-featureNames(genes(cuff))
cufGeneSetObj<-getGenes(cuff,cuff_gene_ids)
ic<-csCluster(cufGeneSetObj,k=5,logMode=T)
clus<-csClusterPlot(ic)
ggsave(file="output.png", plot=clus)
#########################
=================================
The error message we met was as follows.
=================================
Error in pam(n, k, ...) :
NA-values in the dissimilarity matrix not allowed.
Calls: csCluster -> csCluster -> pam
In addition: Warning messages:
1: In sqrt(JSdiv) : NaNs produced
2: In sqrt(JSdiv) : NaNs produced
Execution halted
=================================
We found the data set from our cuffdiff output which makes cummeRbund stop.
If cummeRbund loads the data set, JSdiv becomes a negative value.
So, we suppose that the error might be occurred when JSdiv had a negative
value.
The following is the data set we found.
Data type is data.frame, and the data include two sample groups and two
genes.
The value is FPKM.
=================================
> test_data
group01 group02
gene1 9.04721 3.52402
gene2 10.97890 4.07558
=================================
The error was reproduced when the data set was loaded to JSdiv as follows,
=================================
> test_data.log<-log10(test_data+1)
> JSdist(makeprobs(t(test_data.log)))
gene1
gene2 NaN
Warning messages:
1: In sqrt(JSdiv) : NaNs produced
2: In sqrt(JSdiv) : NaNs produced
=================================
Please let us know how to fix this problem.
Sincerely yours, ytanaka
