Re: [rsem-users] Differential expression analysis on RSEM processed data

[Ning Leng]

Hi Dvir,

You might consider EBSeq
http://www.biostat.wisc.edu/~kendzior/EBSEQ/

RSEM-EBSeq pipeline
http://deweylab.biostat.wisc.edu/rsem/README.html#de

Best,
Ning

On Tue, Mar 26, 2013 at 1:10 AM, Dvir <dvir...@gmail.com> wrote:
> Hello,
>
> Which methods can be applied on RSEM output files to detect differentially
> expressed genes?
> It seems that methods like EdgeR and DESeq require gene level raw count and
> therefore cannot be applied to RSEM outputs.
>
> I have RSEM files of gene level downloaded from TCGA web site.
>
> Many thanks,
> Dvir
>
> --
> You received this message because you are subscribed to the Google Groups
> "RSEM Users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to rsem-users+...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>



--
Ning Leng
University of Wisconsin Madison
Department of Statistics
4720 Medical Sciences Center
1300 University Avenue
Madison, Wisconsin 53706

[Colin Dewey]

Hi Dvir,

In the RSEM output, you should find an "expected count" column. After rounding these values to the nearest integer, you can use them with differential expression packages such as EBSeq, edgeR, and DESeq.

If that column is not available in the TCGA data, you might consider contacting TCGA to have them provide that data.

Best,
Colin

[b...@cs.wisc.edu]

Hi Dvir,

As Ning mentioned, RSEM includes EBSeq and associated scripts to detect DE
genes for you. Also, you should be able to give the gene level expected
counts (rounded to nearest integer) to edgeR or DESeq. Although it is not
ideal, it is what most people do.

Best,
Bo

[Dvir]

Hi and thanks...

 

As I wrote to Colin, my problem is that I need to somehow hook up to specific data files available by TCGA which are in some kind of rsem output format, but I'm not sure if they contain the 'expected count' values.

 

My other option is to download the extremely large sequencing data files and run RSEM and them EBseq locally, but we're talking on hundreds of very large files so this may be challenging, and I'd rather use the already provided Level 3 TCGA output files, even though I'm not sure how to interpret their format.

 

Thanks,

Dvir

[Dvir]

Hi again,

 

It seems like my previous post didn't get through so I'll repost...

 

I have the following files available for me at the TCGA web site, all are supposed to be RSEM output files:

unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_exon.txt
unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_junction.txt
unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_gene.txt
unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_gene_normalized.txt
unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_isoforms.txt
unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_isoforms_normalized.txt

 

For the differential analysis I wish to conduct I have focused on the '..._expression_rsem_gene.txt' files which are the gene-level unnormalized output. This file is of the following format:

 

barcode		gene_id		raw_count		scaled_estimate		transcript_id
TCGA-AN-A03X-01A-21R-A00Z-07		?|100130426		0		0		uc011lsn.1
TCGA-AN-A03X-01A-21R-A00Z-07		?|100133144		12		5.65E-07		uc010unu.1,uc010uoa.1
TCGA-AN-A03X-01A-21R-A00Z-07		?|100134869		0		0		uc002bgz.2,uc002bic.2
TCGA-AN-A03X-01A-21R-A00Z-07		?|10357		123		1.34E-05		uc010zzl.1
TCGA-AN-A03X-01A-21R-A00Z-07		?|10431		1360		6.68E-05		uc001jiu.2,uc010qhg.1
TCGA-AN-A03X-01A-21R-A00Z-07		?|136542		0		0		uc011krn.1
TCGA-AN-A03X-01A-21R-A00Z-07		HLA-A|3105		17614.94		0.00063316		uc003nok.2,uc003nol.2,uc003nom.2,uc003non.2,uc003noo.2,uc010jrq.2,uc010jrr.2,uc010klp.2,uc011dmc.1,uc011dmd.1
TCGA-AN-A03X-01A-21R-A00Z-07		HLA-B|3106		20212.19		0.000956254		uc003ntf.2,uc003ntg.1,uc003nth.2,uc003nti.1,uc010jsm.1,uc010jsn.1,uc010jso.2,uc011dnk.1
TCGA-AN-A03X-01A-21R-A00Z-07		HLA-C|3107		17775.81		0.000668874		uc003nsx.2,uc003nsy.2,uc003nsz.2,uc003nta.2,uc003ntb.2,uc003ntc.1,uc010jsl.2,uc011dnj.1,uc011dnl.1

 

You mentioned that should use the 'expected counts' values, but they only thing similar is the 'raw_count' column in the above file.

 

I wonder if anyone knows this file format and can tell me if I can use the 'raw_count' column for DE analysis (I'm not if this file format is the original RSEM output format or a TCGA format).

 

Many thanks,

Dvir.

 

 

On Tuesday, March 26, 2013 5:31:00 PM UTC+2, leng...@gmail.com wrote:

[Bo Li]

Hi Dvir,

Can you tell me where I can find the RSEM like outputs you mentioned? Then maybe I can give you some suggestions on how to run DE tools on them.

Best,
Bo

[Bo Li]

Hi Dvir,

I think that you are safe to use the "raw_count" column. This is the expected counts we generated. You can see that there are 2 digits after the decimal point, which suggests it is not raw count ( integers) but expected count (real numbers). I'm not sure why they rename the fields in that way...

Best,
Bo

[Dvir]

Thanks Bo.

 

So I will use the raw_count column as input for DESeq/EdgeR after rounding the values. Can I also use them for EBSeq or do I need the entire RSEM output for that?

 

And another question if you will - for clustering analysis, should I use the normalized files ? They have the following format:

 

barcode		gene_id		normalized_count
TCGA-A8-A082-01A-11R-A00Z-07		?|100130426		0
TCGA-A8-A082-01A-11R-A00Z-07		?|100133144		9.372
TCGA-A8-A082-01A-11R-A00Z-07		?|100134869		1.028
TCGA-A8-A082-01A-11R-A00Z-07		?|10357		130.64
TCGA-A8-A082-01A-11R-A00Z-07     Thanks, Dvir		?|10431		1015.2

[Ning Leng]

Hi Dvir,

EBSeq doesn't require integer inputs. So you could use the raw_count matrix.

And I think it makes more sense to use normalized values for clustering analysis.

Best,

Ning

--

You received this message because you are subscribed to the Google Groups "RSEM Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to rsem-users+...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Ning Leng]

Hi Dvir,

I think raw_count is the expected count from rsem.

In the DESCRIPTION.txt in the data folder:

RSEM abundance estimation results in two files, gene and isoform level quantification. More

information regarding the content of these output files can be found on the RSEM website

(http://deweylab.biostat.wisc.edu/rsem/rsem-calculate-expression.html#output).  In short,

the format indicates the feature name in column 1, esimated count in colum 2, scaled

estimate in column 3, and contributing isoforms in column 4 (gene level only). These files

will have the following extensions:

rsem.genes.results

rsem.isoforms.results

I believe the "estimated count" is the expected count.

Best,

Ning

On Fri, Mar 29, 2013 at 1:34 AM, Dvir <dvir...@gmail.com> wrote:

Hi,

 

I use the Data Portal's data matrix to download the relevant RNA-Seq data.

 

 

Using this link -

https://tcga-data.nci.nih.gov/tcga/dataAccessMatrix.htm?mode=ApplyFilter&diseaseType=GBM

I choose

Data Type -> RNASeqV2

Data Level -> Level 3

Availability -> Available

 

Then, in the data matrix, I select a few samples and can build an archive containing the files I mentioned for each sample.

 

Thanks !

Dvir

 

On Tuesday, March 26, 2013 5:31:00 PM UTC+2, leng...@gmail.com wrote:

[b...@cs.wisc.edu]

Hi Dvir,

It appears that they changed the format somehow. But the raw count field
should be the expected count field and scaled_estimate should be tau
values.

Unfortunately, you cannot use 'rsem-generate-data-matrix' to collect the
data you need. But after you have the data matrix, all following steps
should be the same as RSEM's tutorial. In addition, you should be able to
modify 'rsem-generate-data-matrix' to collect the required data from TCGA
files.

Best,
Bo

> Hi and thanks for your help.
>
> I'm not sure how easy it would be to get the TCGA data in a different
> format so I want to be sure the currently available one isn't the one I
> need.
> TCGA data contains the following files for each sample:

>
> unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_exon.txt
>
> unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_junction.txt
>
> unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_gene.txt
>
> unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_gene_normalized.txt
>
> unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_isoforms.txt
>
> unc.edu__IlluminaHiSeq_RNASeqV2__TCGA-A7-A0D9-01A-31R-A056-07__expression_rsem_isoforms_normalized.txt
>

> I focused on the *....rsem_expression_gene.txt* files for the differential
> analysis because they contain a column called 'raw_count' which seems
> closest to the 'expected count' column you mentioned.
>
>
> Here's the format of this file :
>
>
> barcode
>
> gene_id
>
> *raw_count*

> Is this format familiar to you or is it some kind of TCGA format? Do you
> think the raw_count column on the above file format is the one I can use
> instead of expected_count to run DESeq/EdgeR ?
>
> Many thanks,
> Dvir

>
>
>
>
> On Tuesday, March 26, 2013 5:36:03 PM UTC+2, Colin Dewey wrote:
>
>> Hi Dvir,
>>
>> In the RSEM output, you should find an "expected count" column. After
>> rounding these values to the nearest integer, you can use them with
>> differential expression packages such as EBSeq, edgeR, and DESeq.
>>
>> If that column is not available in the TCGA data, you might consider
>> contacting TCGA to have them provide that data.
>>
>> Best,
>> Colin
>>
>> On Mar 26, 2013, at 10:31 AM, Ning Leng <leng...@gmail.com

>> <javascript:>>

>> wrote:
>>
>> > Hi Dvir,
>> >
>> > You might consider EBSeq
>> > http://www.biostat.wisc.edu/~kendzior/EBSEQ/
>> >
>> > RSEM-EBSeq pipeline
>> > http://deweylab.biostat.wisc.edu/rsem/README.html#de
>> >
>> > Best,
>> > Ning
>> >
>> > On Tue, Mar 26, 2013 at 1:10 AM, Dvir <dvir...@gmail.com

>> <javascript:>>

>> wrote:
>> >> Hello,
>> >>
>> >> Which methods can be applied on RSEM output files to detect
>> differentially
>> >> expressed genes?
>> >> It seems that methods like EdgeR and DESeq require gene level raw
>> count
>> and
>> >> therefore cannot be applied to RSEM outputs.
>> >>
>> >> I have RSEM files of gene level downloaded from TCGA web site.
>> >>
>> >> Many thanks,
>> >> Dvir
>> >>
>> >> --
>> >> You received this message because you are subscribed to the Google
>> Groups
>> >> "RSEM Users" group.
>> >> To unsubscribe from this group and stop receiving emails from it,
>> send
>> an

>> >> email to rsem-users+...@googlegroups.com <javascript:>.

>> >> For more options, visit https://groups.google.com/groups/opt_out.
>> >>
>> >>
>> >
>> >
>> >
>> > --
>> > Ning Leng
>> > University of Wisconsin Madison
>> > Department of Statistics
>> > 4720 Medical Sciences Center
>> > 1300 University Avenue
>> > Madison, Wisconsin 53706
>> >
>> > --
>> > You received this message because you are subscribed to the Google
>> Groups "RSEM Users" group.
>> > To unsubscribe from this group and stop receiving emails from it, send

>> an email to rsem-users+...@googlegroups.com <javascript:>.