[Bioc-sig-seq] using DEXSeq for up and down regulated exons

Dario Strbenac D.Strbenac at garvan.org.au
Thu Aug 18 01:00:14 CEST 2011


Because you only have one exon that is expressed by the samples, the two exons can't be tested for different proportions of usage.

---- Original message ----
>Date: Wed, 17 Aug 2011 17:12:03 +0200
>From: bioc-sig-sequencing-bounces at r-project.org (on behalf of Jane Merlevede <jane.merlevede at gmail.com>)
>Subject: Re: [Bioc-sig-seq] using DEXSeq for up and down regulated exons  
>To: bioc-sig-sequencing at r-project.org
>
>Dear Wolfgang, dear Simon,
>
>Thanks for the explanation, I did not have understand well the purpose
>of DEXSeq.
>If you are interested, here are the cases in my dataset where genes
>have more than 1 exon, but testGeneForDEU did not give a result:
>
>*The simplest case:*
>res1[1449:1451,]
>                   geneID exonID dispersion_CR_est   dispersion   pvalue
>EHI_031280:001 EHI_031280   E001        0.25177357 2.517736e-01 0.837492
>EHI_031280:002 EHI_031280   E002        0.03580064 3.580064e-02 0.837492
>EHI_031280:003 EHI_031280   E003                NA 1.000000e+08       NA
>                 padjust
>EHI_031280:001 0.9746982
>EHI_031280:002 0.9746982
>EHI_031280:003        NA
>
>And the read counts per exon are:
>exon_counts[1449:1451,]
>            exon_ID  FKKM1_1 FPKM1_2 FPKM1_3 FPKM2_1 FPKM2_2 FPKM2_3
>1449 EHI_031280:001    9      9      5       33      7    2
>1450 EHI_031280:002  155    192    220      453    133  176
>1451 EHI_031280:003    0      0      0        0      0    0
>
>One exon in the gene has 0 for all FPKM => 10^8 dispersion and no
>result for the test. I am not surprised, contrary to *this second
>case*:
> res1[1227:1228,]
>                   geneID exonID dispersion_CR_est  dispersion pvalue padjust
>EHI_025270:001 EHI_025270   E001                NA 1.00000e+08     NA      NA
>EHI_025270:002 EHI_025270   E002                NA 2.03465e-02     NA      NA
>
>And the read counts per exon are:
> exon_counts[1227:1228,]
>            exon_ID  FKKM1_1 FPKM1_2 FPKM1_3 FPKM2_1 FPKM2_2 FPKM2_3
>1227 EHI_025270:001     0      0      0        0      0    0
>1228 EHI_025270:002 11900   3012  12136    10971   7387 5923
>
>Why is there no result for the second exon?
>
>Thanks,
>Jane
>
>
>"Dear Jane,
>
>in the case of genes (or better term: loci) with only one exon, DEXSeq's
>testGeneForDEU does not make sense - it is looking for exons within the
>locus that behave differently from the average of all the others: there
>are no others! Hence the parameters are unidentifiable, and the software
>should not even look at them. We'll robustify the software for this case
>in future versions (note that the package is still actively being
>developed).
>
>The case of the NA values that are not explained by there only being one
>exon is more interesting. Can you send us the code to reproduce your
>problem (this also requires the data, which you can send offline, and in
>which you can for instance anonymize the gene and sample annotations).
>
>	Best wishes
>	Wolfgang
>
>Hi Jane
>
>On 08/08/2011 02:45 PM, Wolfgang Huber wrote:
>>* in the case of genes (or better term: loci) with only one exon, DEXSeq's
>*>* testGeneForDEU does not make sense - it is looking for exons within the
>*>* locus that behave differently from the average of all the others: there
>*>* are no others! Hence the parameters are unidentifiable, and the software
>*>* should not even look at them.
>*
>Actually, it doesn't. This is why the p value is NA. Also note the
>column 'testable' in the fData data frame: it is set to FALSE in such
>cases to indicate that a test was not attempted. You will also find
>testable=FALSE occasionally in genes with more than one exons, namely
>whenever all exons or all exons but one have zero counts in all samples.
>
>Cheers
>   Simon"
>
>
>
>2011/7/29 Jane Merlevede <jane.merlevede at gmail.com>
>
>> Hello,
>>
>> I am using DEXseq for differential analysis. I have posted some e-mails
>> about it already on this list, but I have more questions !
>> My dataset has 2 conditions with 3 biological replicates per condition:
>>
>> pData(ecs)
>>         sizeFactor condition replicate       type
>> Raman_1  0.9638822    nonvir         1 paired-end
>> HM1_1    1.4000715       vir         1 paired-end
>> Raman_2  1.0314049    nonvir         2 paired-end
>> Raman_3  1.0734133    nonvir         3 paired-end
>> HM1_2    0.7801269       vir         2 paired-end
>> HM1_3    0.9488409       vir         3 paired-end
>>
>>
>> I found the number of differentially expressed exons. I would like to know
>> also which are down and which are up regulated. This information is given by
>> foldChange, which is not provided in DEXSeq (res1).
>>
>> res1 <- DEUresultTable(ecs)
>> head(res1)
>>                    geneID exonID dispersion_CR_est dispersion    pvalue
>> EHI_000010:001 EHI_000010   E001                NA 0.02377712        NA
>> EHI_000130:001 EHI_000130   E001        0.07466945 0.07466945 0.4867354
>> EHI_000130:002 EHI_000130   E002        0.12897281 0.12897281 0.9427953
>> EHI_000130:003 EHI_000130   E003        0.01960222 0.02031871 0.8432791
>> EHI_000130:004 EHI_000130   E004        0.06733720 0.06733720 0.7042977
>> EHI_000240:001 EHI_000240   E001        0.04375400 0.04375400 0.7493987
>>                  padjust
>> EHI_000010:001        NA
>> EHI_000130:001 0.8639624
>> EHI_000130:002 0.9927663
>> EHI_000130:003 0.9756197
>> EHI_000130:004 0.9503086
>> EHI_000240:001 0.9611397
>>
>> So I would like to know, if it could be possible to have baseMean,
>> baseMeanA, baseMeanB, foldChange, log2FoldCHange, resVarA and resVarB as in
>> DESeq?
>>
>> Since this does not seem to be implemented in DEXSeq, I should start with
>> the read counts per exon table to get it, but I can't find it neither. I
>> think it is use in the estimateSizeFactors function, but I don't manage to
>> access it. It's a pity, I will have to reconstruct the table...
>>
>>
>> Thank you for your help,
>> Jane Merlev�de
>>
>
>	[[alternative HTML version deleted]]
>
>________________
>_______________________________________________
>Bioc-sig-sequencing mailing list
>Bioc-sig-sequencing at r-project.org
>https://stat.ethz.ch/mailman/listinfo/bioc-sig-sequencing


--------------------------------------
Dario Strbenac
Research Assistant
Cancer Epigenetics
Garvan Institute of Medical Research
Darlinghurst NSW 2010
Australia


More information about the Bioc-sig-sequencing mailing list