Error message in DESeq (estimateVaranceFunctions, preplot.locfit.raw)

[aquila]

Hello all, I am receiving an error message in DESeq which I do not understand. I have data from 14 samples (7 samples for condition A and 7 samples for condition B). If I run DESeq on all samples, it works without error messages. However, if I just choose a subset of the samples (3 samples for condition A and 3 samples for condition B), it may or may not work, depending on the subset.

When trying to run estimateVarianceFunctions(cds), I get the following error message and no variance functions are added:

Error in preplot.locfit.raw(object, newdata, where, what, band) :  
  NA/NaN/Inf in foreign function call (arg 2)

I have not been able to figure out where the function encounters these NA/NaN/Inf values and what could be done to make it work. Does anyone have suggestions?

In addition, the size factors are all estimated to 1. This seems to cause the following warnings:

1: In lfproc(x, y, weights = weights, cens = cens, base = base, geth = geth,  :  
  compparcomp: perfect fit
2: In lfproc(x, y, weights = weights, cens = cens, base = base, geth = geth,  :  
  procv: parameters out of bounds

It does not appear to me that this is the culprit that causes the variance function estimation to fail, but I may be wrong.

Any help is appreciated.

Regards,
aquila

[NicoBxl]

have you any NA value in your count table ?

if there are some NA values, you can change them to 0 and retry DESeq .

Otherwhise Simon Anders can help you

[aquila]

Hi NicoBxl,
Thanks for your reply. No, I do not have NA values in my count table.
The strange thing is that the analysis runs without error if I use all my samples. The error occurs only if I leave out some of the samples.

[NicoBxl]

can you copy paste your commands

[aquila]

Hi, thank you for your continued help. The issue has been resolved. The reason for the error message was a problem with my input data, not with my DESeq usage. The error seems to occur if there are no entries where the count is 0.