Question

DESeq estimateSizeFactors() is returning error "Error: all(!is.na(value)) is not TRUE"

0

Entering edit mode

abhishek.yedugondla • 0

@abhishekyedugondla-14108

Last seen 6.5 years ago

So upstream seems to be working well, I have run into an issue where the estimatesizefactors is returning a strange error. Input is:

dds <- dds[ rowSums(counts(dds)) > 1, ]

cts <- counts(dds)

geoMeans <- apply(cts, 1, function(row) if (all(row == 0)) 0 else exp(mean(log(row[row != 0]))))

dds <- estimateSizeFactors(dds, geoMeans=geoMeans)

This returns the error

> Error: all(!is.na(value)) is not TRUE

This results in a downstream error of:

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

Ran the following to troubleshoot:

traceback()
table(rowSums(counts(dds) == 0 ))
table(colSums(counts(dds) == 0))

Resulted in :

> traceback() 6: stop("every gene contains at least one zero, cannot compute log geometric means") 5: estimateSizeFactorsForMatrix(counts(object), locfunc = locfunc, geoMeans = geoMeans, controlGenes = controlGenes) 4: .local(object, ...) 3: estimateSizeFactors(object) 2: estimateSizeFactors(object) 1: rlogTransformation(dds) > table(rowSums(counts(dds) == 0 ))

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 10933 926 705 560 488 450 417 396 384 402 419 419 467 469 534 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 582 723 908 799 567 622 539 588 546 550 651 646 702 727 833 31 32 33 34 35 713 348 94 7 1 > table(colSums(counts(dds) == 0))

6372 6503 6575 6699 6757 6847 6923 6994 7005 7009 7024 7105 7203 7307 7377 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 8246 8280 8447 8557 8804 8907 9655 9783 10338 10345 10358 10485 10504 10524 10606 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 10699 10700 11427 11544 11580 29115 1 1 1 1 1 1

Don't See any zero numbers in the tables so I am unsure as to what is causing this error. The fix of removing all values less than 1 should have fixed it no? Please advise.

deseq2 error R estimatesizefactors zero • 5.5k views

ADD COMMENT • link updated 2.0 years ago by LW • 0 • written 6.5 years ago by abhishek.yedugondla • 0

score 0 · Answer 1 · 2017-10-04

0

Entering edit mode

Michael Love 41k

@mikelove

Last seen 5 hours ago

United States

Can you try estimateSizeFactors(dds, type="poscount")?

This was implemented recently, although proposed a while ago on phyloseq/DESeq2 forums, and hopefully avoids users having to do their own estimation of the geometric means, etc.

ADD COMMENT • link 6.5 years ago Michael Love 41k

0

Entering edit mode

Hi,

I got this error, and tried estimateSizeFactors(dds, type="poscount"), but that error is still there. If I add 1 to each cell of the count table, the error is gone. Is it valid to just add 1?

Thanks! Leran

ADD REPLY • link 2.0 years ago LW • 0

0

Entering edit mode

I wouldn’t add one. If you have no genes without 0s, it may indicate you should be using a different scaling factor procedure. Is this single cell or microbiome?

ADD REPLY • link 2.0 years ago Michael Love 41k

0

Entering edit mode

this is microbiome 16S data

ADD REPLY • link 2.0 years ago LW • 0

0

Entering edit mode

I don't think DESeq2 is appropriate for microbiome (there are a couple of threads on this).

If the size factor doesn't work, neither will the downstream steps.

ADD REPLY • link 2.0 years ago Michael Love 41k

0

Entering edit mode

Thanks Michael, I double checked and found that there are 4 samples are with 0 taxa. When I remove them, the software works fine!

Leran

ADD REPLY • link 2.0 years ago LW • 0

score 0 · Answer 2 · 2017-12-04

0

Entering edit mode

fanli.gcb • 0

@fanligcb-10919

Last seen 6.4 years ago

Los Angeles, CA

Check to make sure all of the column sums of your count matrix are non-zero. I encountered this error and eventually figured out that it was due to a typo in my sample sheet (a misnamed sample ended up with all zero counts).

ADD COMMENT • link 6.4 years ago fanli.gcb • 0