kmeans clustering in R

[Galgarad]

Hi guys,

I'm new to R and I'm really close to throwing my cpu in a lake...

I just can't seem to get the kmeans script to run on my data, using R.

I really hope you can help me on this.. I have a tab delimited text file with the first column being the nam of the genes, and the two following columns are expression data.

Ex:

TPM3 0.014 0.683
TFF1 4.5 3.567
....


While in R, i do the following:
> data<-read.table("file path to the data", row.names=1)
> datam<-as.matrix(data)
>k<-kmeans(datam, 3)


and then I have that *&/%/$ error :
Error in do_one(nmeth) : NA/NaN/Inf in foreign function call (arg 1)
In addition: Warning message:
In kmeans(datam, 3) : NAs introduced by coercion


I've been searching the web for the anwser a lot, and I couldn't find how to fix it. I tried loading a .csv instead of a .txt, but nothing works. From what I have read, it seems that the problem is that there are some blank spaces in my file. But even if I do the
> datak<-na.omit(datam)
> k<-kmeans(datak, 3)

I get the same error..

Please help me!!!

Phil

[Bukowski]

I don't think it's blank spaces in the data - R would most likely spit a scan error about the lack of elements in the input before you even got to passing it to kmeans.

I think you most likely have non-numeric data in one of your numeric input columns. Sanity check the input file first.

See:

> data
V2 V3
gene1 1 2
gene2 3 4
gene7 9 0
gene3 4 5
gene4 5 6
> k<-kmeans(data, 3)
> k
K-means clustering with 3 clusters of sizes 3, 1, 1

vs

> data
V2 V3
gene1 1 2
gene2 3 4
gene7 9 0
gene3 4 5
gene4 5 6
gene9 F 10
> k<-kmeans(data, 3)
Error in do_one(nmeth) : NA/NaN/Inf in foreign function call (arg 1)
In addition: Warning message:
In do_one(nmeth) : NAs introduced by coercion

[Galgarad]

You were right.

I had a 3,9 E + 10 somewhere.

I thought I checked the file thouroughly, but it seems like I didn't

Thanks a lot!!!

[jp.]

hi all
I got the same error which is "Error in hclustfun(distfun(if (symm) x else t(x))) :
NA/NaN/Inf in foreign function call (arg 11)".
I have multiple missing values and I have entered NA. If there few "NA", it goes well and makes heatmap with white (no color) of NA.
But if there are many NA then only it gives error...
My commands are:
> t<- as.matrix(read.table("tmp.txt", sep="\t",header=T, na.strings = "NA", as.is=T,row.names=1))
> heatmap.2(t, trace="none", ylab="Gene_Name", xlab="Sample_Name", margin=c(7,20), density.info="histogram",na.color=NA, keysize=0.8, densadj=0.025, scale="row", cexRow=1, cexCol=1,dendrogram="both",Rowv=T,Colv=T,key=T,hclustfun=hclust,col = greenred(100))

Can anybody give me solution ? I want to make heatmap with multiple or numerous NA values.
Thank you
jp.