Hi, I am wondering why Cuffdiff suddenly gives many more significant DE genes? I have used same input data and now get approx 5x more significant genes, settings is same with the exception that you now included library normalization and dispersion estimation. See below for parameters. I have rerun several analysis, also with more samples in each group, all give much more significant genes. Why? Also, will replicate information soon be included in output files? Kind regards, Johanna Tool: Cuffdiff Tool: Cuffdiff Name: Cuffdiff on data 225, data 236, and others: splicing differential expression testing Name: Cuffdiff on data 225, data 236, and others: splicing differential expression testing Created: 4-Apr-13 Created: 21-Aug-13 Filesize: 10.3 MB Filesize: 10.2 MB Dbkey: hg19 Dbkey: hg19 Format: tabular Format: tabular Galaxy Tool Version: 0.0.5 Galaxy Tool Version: 0.0.5 Tool Version: Tool Version: Tool Standard Output: stdout<https://main.g2.bx.psu.edu/datasets/bbd44e69cb8906b5355e4a035e92cd84/stdout> Tool Standard Output: stdout<https://main.g2.bx.psu.edu/datasets/bbd44e69cb8906b524bcf7e585f495b1/stdout> Tool Standard Error: stderr<https://main.g2.bx.psu.edu/datasets/bbd44e69cb8906b5355e4a035e92cd84/stderr> Tool Standard Error: stderr<https://main.g2.bx.psu.edu/datasets/bbd44e69cb8906b524bcf7e585f495b1/stderr> Tool Exit Code: 0 Tool Exit Code: 0 API ID: bbd44e69cb8906b5355e4a035e92cd84 API ID: bbd44e69cb8906b524bcf7e585f495b1 Input Parameter Value Note for rerun Input Parameter Value Transcripts 261: Cuffmerge on data 258, data 135, and others: merged transcripts Transcripts 261: Cuffmerge on data 258, data 135, and others: merged transcripts Perform replicate analysis Yes Perform replicate analysis Yes Group name s202 Group name s202 Add file 194: Galaxy883-[MarkDups_Dupes_Marked_882_202.bam].bam Add file 194: Galaxy883-[MarkDups_Dupes_Marked_882_202.bam].bam Group name Ctrls Group name Ctrls Add file 236: MarkDups_Dupes Marked216.bam Add file 236: MarkDups_Dupes Marked216.bam Add file 225: MarkDups_Dupes Marked206.bam Add file 225: MarkDups_Dupes Marked206.bam Library normalization method not used (parameter was added after this job was run) Library normalization method geometric Dispersion estimation method not used (parameter was added after this job was run) Dispersion estimation method pooled False Discovery Rate 0.05 False Discovery Rate 0.05 Min Alignment Count 2 Min Alignment Count 2 Perform quartile normalization No Perform quartile normalization No Use multi-read correct Yes Use multi-read correct Yes Perform Bias Correction Yes Perform Bias Correction Yes Reference sequence data cached Reference sequence data cached Set Additional Parameters? (not recommended) Yes Set Additional Parameters? (not recommended) Yes Average Fragment Length 200 Average Fragment Length 200 Fragment Length Standard Deviation 80 Fragment Length Standard Deviation 80 ...................................................................................................................................................... Johanna Sandgren, PhD Department of Oncology-Pathology CCK, Karolinska Institutet SE-171 76 Stockholm, Sweden +46-8-517 721 35 (office), +46-8- 321047(fax), +46-708 388476 (mobile)
I am wondering why Cuffdiff suddenly gives many more significant DE genes?
Cuffdiff was recently updated to version 2.1.0; this update likely explains the different results that you see.
I have rerun several analysis, also with more samples in each group, all give much more significant genes. Why?
More samples = more power to accurately estimate expression levels = more DE genes.
Also, will replicate information soon be included in output files?
Replicate data will be available when we next update our server. This should occur in about 3 weeks. Best, J.
Thanks for quick reply. Where can I see which version are being used? It does not say in the attached (in my first e-mail) info-view after the run.. What does Cuffdiff (version 0.0.5) mean then? What version was it before? It is a great difference in number of DE genes.. I mostly meant it was the same now in August with more DE genes for all different type of analysis that I have previously done (consistent new results with no settings changed), I really do not see more DE genes when 5 vs 6 samples compared to 1 vs 2 samples. But that might be due to more biological reasons. I look forward to the update, will that mean another version of Cuffdiff again? Kind regards, Johanna From: Jeremy Goecks [mailto:jeremy.goecks@emory.edu] Sent: Thursday, August 22, 2013 8:04 PM To: Johanna Sandgren Cc: galaxy-user@lists.bx.psu.edu Subject: Re: [galaxy-user] Cuffdiff changes I am wondering why Cuffdiff suddenly gives many more significant DE genes? Cuffdiff was recently updated to version 2.1.0; this update likely explains the different results that you see. I have rerun several analysis, also with more samples in each group, all give much more significant genes. Why? More samples = more power to accurately estimate expression levels = more DE genes. Also, will replicate information soon be included in output files? Replicate data will be available when we next update our server. This should occur in about 3 weeks. Best, J.
Where can I see which version are being used?
You can see both the Galaxy tool version and the Cuffdiff tool version (when available) by clicking on the 'view details' icon (the 'i' at the bottom of an expanded dataset). Right now the Cuffdiff version is not displayed, but that will change when our server is updated.
What does Cuffdiff(version 0.0.5) mean then?
That is the version of the Galaxy wrapper; the wrapper provides the interface between Cuffdiff and Galaxy.
What version was it before?
I think Cuffdiff version was 1.3.1 previously.
I look forward to the update, will that mean another version of Cuffdiff again?
The wrapper will be updated but not Cuffdiff itself. J.
participants (2)
-
Jeremy Goecks
-
Johanna Sandgren