NOISeq with fpkm values

[samchen]

I met the same problem.
There are some non-numericl values of your fpkms/rpkms.
NOISeq doesn't recognise the values like "2.24343e-2" because it can't be translated into "0.0224343". However, you can use "2.24343E-2".
Try it. Finally, I got the results.

[11xinqi]

Good morning all!

Recently I've been doing comparisons of differential expression calling of several different packages. Really, I've been comparing Cufflinks against other count-based approaches. I also have used NOISeq with raw counts but it says in documentation that it can do rpkm. Sooo, I wanted to take the fpkm values from Cufflinks and use them in NOISeq. Seemed simple enough...

However, I keep getting this: "Error in rowSums(as.matrix(datos1)) : 'x' must be numeric" and "Error in round(datos1, 100) : Non-numeric argument to mathematical function".

Now, I've specifically imported my file and forced it to be a numeric data matrix so now I'm lost as to what else to try. Has anyone else had this problem? Is NOISeq not capable of using these fpkm values?

Thanks in advance!

Hi,

I am doing the same job. I have 2 samples and each sample has 3 biological replicates. Would you give me a pipeline about how to make NOISeq work with FPKM ? Thank you very much

[NitaC]

I think I agree sdriscoll.

One question though: With NOISeq, do you keep k=0 when using FPKMs or do you use k=.5 (or some other small number)?

[sdriscoll]

I actually think the NOISeq method performs better with FPKMs.

[NitaC]

*Problem Solved*

In case anyone ever comes across this, I wanted to note that the problem was with R not NOISeq. I had an issue with one column being treated as factors rather than numeric data. As I'm also a newb to R, this was an unexpected issue. However, now that the data is indeed numeric, NOISeq is working just fine.

[NitaC]

Good morning all!

Recently I've been doing comparisons of differential expression calling of several different packages. Really, I've been comparing Cufflinks against other count-based approaches. I also have used NOISeq with raw counts but it says in documentation that it can do rpkm. Sooo, I wanted to take the fpkm values from Cufflinks and use them in NOISeq. Seemed simple enough...

However, I keep getting this: "Error in rowSums(as.matrix(datos1)) : 'x' must be numeric" and "Error in round(datos1, 100) : Non-numeric argument to mathematical function".

Now, I've specifically imported my file and forced it to be a numeric data matrix so now I'm lost as to what else to try. Has anyone else had this problem? Is NOISeq not capable of using these fpkm values?

Thanks in advance!