DEXSeq - able to analyze paired (matched) data?

[spabinger]

Hi everyone,

does anybody know if DEXSeq is able to analyze paired (matched) data - e.g.: same patient before and after treatment?

If not, is any software capable of handling the analysis of differential exone usage in RNA-Seq paired data (don't confuse it with paired end)?

Thanks in advance,
Stephan

[areyes]

Hi Stephan,

Yes it is possible by changing the formula at the time of estimating dispersions and testing for differential exon usage. For more information check the vignette

http://bioconductor.org/packages/2.9/bioc/vignettes/DEXSeq/inst/doc/DEXSeq.pdf

Alejandro

[spabinger]

Hi Alejandro,

thanks a lot for your reply. I will have a look at it.

Best
Stephan

[townway]

how about several treatments and also samples without replicates, does DEXSeq handle as well? Thank you

Hi Stephan,

Yes it is possible by changing the formula at the time of estimating dispersions and testing for differential exon usage. For more information check the vignette

http://bioconductor.org/packages/2.9/bioc/vignettes/DEXSeq/inst/doc/DEXSeq.pdf

Alejandro

[areyes]

About several treatments, you can add them to the design data frame at the time of creating the ExonCountSet object. At the time of testing the analysis would tell you which exons are being affected due to changes on the treatment, nevertheless you would not know immediately which one the treatment that is causing the differential exon usage, then probably make pairwise comparisons. You could also adapt the formulas to test for the changes being introduced due to a certain treatment.

About the samples with no replicates, DEXSeq shares the same "motivation" as DESeq or edgeR. Meaning, use biological replicates to estimate biological variation using a negative binomial distribution. Though, you could use DEXSeq: the dispersion estimates will tend to 0 and you will have a lot of false positive results, but it does NOT really makes sense to make these kinds of analysis without biological replicates and a proper estimation of variation. Simon Anders and others have discussed lots about this in this forum!

[emilyjia2000]

I am new to DEXseq, I just wonder if I don't have replicates and have only one condition, is it possible to use DEXseq to count exon level reads?
Thanks,
emily

[Simon Anders]

I am new to DEXseq, I just wonder if I don't have replicates and have only one condition, is it possible to use DEXseq to count exon level reads?

DEXSeq is a tool to test for differential exon usgae, so using it with just a single sample is rather pointless. DEXSeq does come with a Python script to count reads for exons. However, I wonder what use the script's output might be for you.

[emilyjia2000]

Hi Simon,

THanks for the reply. I searched the tools that would help to generate exon expression level reads. for example, HTseq, Rsamtools, DEseq and DEXseq, and I am so confused and don't know which one I should choose to use in the analysis.
I have 20 RNA-seq samples (patient data), no replicates on only one condition, but we would like to know the exon level reads on each sample. Do you have any recommendation for the tools?

[Simon Anders]

There are many tools which take a list of genomic intervals (e.g., a list of exon positions) and count how many reads fall into each of the intervals. DEXSeq comes with such a script, GenomicRanges contains function for it, BEDtools might work etc.

I only really cannot fathom why you would want such an analysis. You will end up with a huge table with 20 columns and a few hundred thousand rows. Clearly, you cannot inspect all these rows by simply looking through them, as, I hope, you realize.

Hence you must have some idea what your next step will be, or what you hope to find in the data. The people in this forum might be much more able and willing to help you if you explained what your next planned step will be.

[Lisa]

DESeq in Galaxy

I am wondering if it is possible to put DESeq in Galaxy? Thanks a lot! Lisa

[emilyjia2000]

Hello Simon,

Thanks for the suggestion, we would like to see transcript level, junction reads as well, not just exon level reads. But we are sort of stuck at the exon level reads. We used tophat to do the mapping.
Now I used HTseq-count to do the exon reads. it generated the error message when running it:
Message:
/usr/local/python-2.7/bin/htseq-count: Error: Please provide two arguments.
Call with '-h' to get usage information.

I am not very clear what the 2 arguments mentioned in the Error message?
command line to use:
htseq-count –t exonic_part –s no i ID -m intersection-nonempty myfile.sam Homo_sapiens.GRCh37.65_chr.gff > exoncount.txt

Do I have to sort/reorder sam files before working on htseq-count?

THanks,

[jbrwn]

Hello Simon,

Thanks for the suggestion, we would like to see transcript level, junction reads as well, not just exon level reads. But we are sort of stuck at the exon level reads. We used tophat to do the mapping.
Now I used HTseq-count to do the exon reads. it generated the error message when running it:
Message:
/usr/local/python-2.7/bin/htseq-count: Error: Please provide two arguments.
Call with '-h' to get usage information.

I am not very clear what the 2 arguments mentioned in the Error message?
command line to use:
htseq-count –t exonic_part –s no i ID -m intersection-nonempty myfile.sam Homo_sapiens.GRCh37.65_chr.gff > exoncount.txt

Do I have to sort/reorder sam files before working on htseq-count?

THanks,

the script doesn't like that you don't have a dash (-) before the "i".

[emilyjia2000]

Thanks Jbrwn,
after I erased "i ID" from the command, same error message.

[jbrwn]

Thanks Jbrwn,
after I erased "i ID" from the command, same error message.

bummer.

for exon counting i have reverted to using there supplied script within the dexseq package (dexseq_count.py) which depends on htseq.

i assume you already have the package and used one script to write the flattened gff file. the scripts are located in:
DEXSeq/inst/python_scripts

the options are outlined in the script and account for paired-end reads, strandedness, and quality.