Question

fourCseq - error getZscores

0

Entering edit mode

samuel collombet ▴ 140

@samuel-collombet-6574

Last seen 6.9 years ago

France

Hi,

When I try to call the function getZSCores on my 4C data (mouse) I got the following error:

> data_fcf <- getZScores(data_fc,minCount=10,fitFun=distFitMonotone)
[1] "VP1"
gene-wise dispersion estimates
mean-dispersion relationship
NOTE: fitType='parametric', but the dispersion trend was not well captured by the
function: y = a/x + b, and a local regression fit was automatically substituted.
specify fitType='local' or 'mean' to avoid this message next time.
final dispersion estimates
Error in getZScores(data_fc, minCount = 10, fitFun = distFitMonotone) :
Failed to estimate the parameters of the Variance stabilizing transformation.

Looking at the code it seems due to the fact that parametric fit did not work so he stopped after that.
Is it not possible to go further in the analysis with a local fit?

fourcseq • 2.5k views

ADD COMMENT • link updated 8.3 years ago by felix.klein ▴ 150 • written 8.4 years ago by samuel collombet ▴ 140

score 0 · Answer 1 · 2015-11-21

0

Entering edit mode

Wolfgang Huber ★ 13k

@wolfgang-huber-3550

Last seen 11 days ago

EMBL European Molecular Biology Laborat…

Hi Samuel

thank you for your interest in this package and for reporting the problem. This should be fixable. Can you send a complete (minimal) code example and your sessionInfo? (Plus perhaps also the required input data, privately).

Kind regards

Wolfgang

ADD COMMENT • link 8.4 years ago Wolfgang Huber ★ 13k

0

Entering edit mode

Hi Wolfgang, Im getting a similar error: Error in getZScores(data, removeZeros= TRUEminCount = 30, fitFun = distFitMonotoneSymmetric) : Failed to estimate the parameters of the Variance stabilizing transformation However I only get it while making certain comparisons between my experimental groups. I have 4 groups and I have been able to analyze them between some of the conditions (the four groups) but in some comparisons I get that error. What could be happening? Best regards! Fernando

ADD REPLY • link 5.3 years ago ferbecneu • 0

score 0 · Answer 2 · 2015-11-21

0

Entering edit mode

felix.klein ▴ 150

@felixklein-6900

Last seen 5.8 years ago

Germany

Hi Samuel,

currently the analysis does not work with a local fit. Maybe you can adjust the threshold in the function if you have noisy data.

How do the scatter plots of your replicates look? This should help to find a suitable threshold.

Best regards,

Felix

ADD COMMENT • link 8.4 years ago felix.klein ▴ 150

0

Entering edit mode

Hi Felix,

I am running into the same error.

I have no replicate for the sample (since it was a trial) and that could be the cause of error. Is it possible to use the method when there are no replicates?

I have another issue when running the example (without including 2R). It gives the error

Error in estimateSizeFactorsForMatrix(counts(object), locfunc = locfunc, :
every gene contains at least one zero, cannot compute log geometric means.

Best Regards,

Shruti

ADD REPLY • link 8.3 years ago sinhashruti • 0

score 0 · Answer 3 · 2015-12-21

0

Entering edit mode

felix.klein ▴ 150

@felixklein-6900

Last seen 5.8 years ago

Germany

Hello Shruti,

you need at least two measurements for a viewpoint to estimate the dispersions. Ideally you have replicates from different conditions.

In the second error it seems like most of your counts are 0 or at least one replicate. Check this by looking at the counts table of your objects:

head(counts(fc), 25)

Best regards,

Felix

ADD COMMENT • link 8.3 years ago felix.klein ▴ 150

0

Entering edit mode

Hi Felix,

Thank you for the reply. I was thinking since now DESeq can be used on data without replicate, probably there would be some feature to do the same for FourCSeq.

For the second error, below is the result from head(counts(fc),25). Please note that the data is from the example data in the package but excluding the 2R. I wanted to check if it works for single chromosome or not. Is it that it needs a specific number of rows before calculating?

> head(counts(fc), 25)

testdata_WE_68h_1 testdata_WE_68h_2 testdata_MESO_68h_1 testdata_MESO_68h_2 testdata_WE_34h_1 testdata_WE_34h_2
[1,] 0 13 4 0 0 0
[2,] 6 0 0 2 8 0

Best Regards

Shruti

ADD REPLY • link 8.3 years ago sinhashruti • 0

score 0 · Answer 4 · 2016-01-09

Hi Shruti,

the testdata in the package is only included to provide an example on how the fragment reference is generated and how it is counted. Therefore you get errors when running the getZScore function.

If you run it on your own data make sure, that there is not a replicate that only contains 0 counts. If there is, remove it from your analysis.

Best reagards,

Felix