DESeq: help with analysis
2
0
Entering edit mode
@abhishek-pratap-4927
Last seen 8.5 years ago
United States
Hi Everyone Given the study design( see below) I am somewhat surprised NOT to see any genes being differentially expressed with p.adj < .05 or .10. I would like the help from folks on this mailing list to introspect the analysis and figure out if we overlooked something or data Study Design: 4 cases v/s 4 control ( all replicates are biological) One sample in the case had very low read counts but I thought that should not make a lot of difference in case of DESeq. I did however try to re-run the analysis taking that out but no change as far as the #diff exp genes. Analysis done with : DESeq_1.12.0 > sizeFactors(cds)TRA00010811 TRA00010812 TRA00010816 TRA00010813 TRA00010814 TRA00010817 TRA00010818 0.9834455 1.3122005 0.7464998 1.3739400 1.6820479 0.5693563 0.7763204 metadata condition libType TRA00010811 ablated paired-end TRA00010812 ablated paired-end TRA00010816 control paired-end TRA00010813 ablated paired-end TRA00010814 ablated paired-end TRA00010817 control paired-end TRA00010818 control paired-end Attached: dispersion plot and MA plot Please let me know if you need anything more from my end to help me dig a bit deeper and understand what might be going on here. Thanks a lot, -Abhi -------------- next part -------------- A non-text attachment was scrubbed... Name: MAplot.png Type: image/png Size: 23914 bytes Desc: not available URL: <https: stat.ethz.ch="" pipermail="" bioconductor="" attachments="" 20131015="" 5b5f6e58="" attachment.png=""> -------------- next part -------------- A non-text attachment was scrubbed... Name: deseq_dispersion_plot.png Type: image/png Size: 24178 bytes Desc: not available URL: <https: stat.ethz.ch="" pipermail="" bioconductor="" attachments="" 20131015="" 5b5f6e58="" attachment-0001.png="">
DESeq DESeq • 1.3k views
ADD COMMENT •
0
Entering edit mode
@steve-lianoglou-2771
Last seen 21 months ago
United States
Hi, On Tue, Oct 15, 2013 at 10:50 AM, Abhishek Pratap <abhishek.vit at="" gmail.com=""> wrote: > Hi Everyone > > Given the study design( see below) I am somewhat surprised NOT to see any > genes being differentially expressed with p.adj < .05 or .10. I would like > the help from folks on this mailing list to introspect the analysis and > figure out if we overlooked something or data Providing the code you used for this (not just the plots) would be helpful, but barring that I'd suggest: (1) Redo your analysis using DESeq2. It uses a less conservative over-dispersion estimate per gene, therefore providing more power to detect differential expression. (2) Explore your data to see if your replicates cluster together, etc. Take some guidance from the "Data quality assessment by sample clustering and visualization" of the DESeq2 vignette. Optional: you may want to do this analysis with the (looming) bioc-2.13 version of the packages, but if that's too much of a burden, use the current release. HTH, -steve -- Steve Lianoglou Computational Biologist Bioinformatics and Computational Biology Genentech
ADD COMMENT •
0
Entering edit mode
@lorena-pantano-6001
Last seen 6 months ago
Boston
Hi, this could happen if your model doesn't explain all your design. There is good example in the tutorial about plotting pvalues to see if you get the expected distribution. Sometime I get the same, and in some cases, although FDR is not so low, if I do a heatmap of the top genes, I get my data clustered into the groups I have in my model (cases versus control for instance). So, sometimes there is signal even if FDR is higher than 0.1. But you need to be very careful and really look your data using other visual representation as others have suggested. cheers Lo On Tue, Oct 15, 2013 at 7:50 PM, Abhishek Pratap <abhishek.vit@gmail.com>wrote: > Hi Everyone > > Given the study design( see below) I am somewhat surprised NOT to see any > genes being differentially expressed with p.adj < .05 or .10. I would like > the help from folks on this mailing list to introspect the analysis and > figure out if we overlooked something or data > > > Study Design: > 4 cases v/s 4 control ( all replicates are biological) > One sample in the case had very low read counts but I thought that should > not make a lot of difference in case of DESeq. I did however try to re-run > the analysis taking that out but no change as far as the #diff exp genes. > > Analysis done with : DESeq_1.12.0 > > > sizeFactors(cds)TRA00010811 TRA00010812 TRA00010816 TRA00010813 > TRA00010814 TRA00010817 TRA00010818 > 0.9834455 1.3122005 0.7464998 1.3739400 1.6820479 > 0.5693563 0.7763204 > > > > metadata > > > condition libType > TRA00010811 ablated paired-end > TRA00010812 ablated paired-end > TRA00010816 control paired-end > TRA00010813 ablated paired-end > TRA00010814 ablated paired-end > TRA00010817 control paired-end > TRA00010818 control paired-end > > > > Attached: > dispersion plot and MA plot > > Please let me know if you need anything more from my end to help me dig a > bit deeper and understand what might be going on here. > > Thanks a lot, > -Abhi > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]
ADD COMMENT •
0
Entering edit mode
Thanks Lorena and Steve. Using the DeSeq2 I was able to get a list of genes after FDR adjustment. We do some expected functional enrichment there which is great. I would be curious to know the changes in DESeq2 but I guess the paper is not yet out. Cheers! -Abhi On Tue, Oct 15, 2013 at 1:33 PM, Lorena Pantano <lorena.pantano@gmail.com>wrote: > Hi, > > this could happen if your model doesn't explain all your design. There is > good example in the tutorial about plotting pvalues to see if you get the > expected distribution. > > Sometime I get the same, and in some cases, although FDR is not so low, if > I do a heatmap of the top genes, I get my data clustered into the groups I > have in my model (cases versus control for instance). So, sometimes there > is signal even if FDR is higher than 0.1. But you need to be very careful > and really look your data using other visual representation as others have > suggested. > > cheers > > > > > Lo > > > On Tue, Oct 15, 2013 at 7:50 PM, Abhishek Pratap <abhishek.vit@gmail.com>wrote: > >> Hi Everyone >> >> Given the study design( see below) I am somewhat surprised NOT to see any >> genes being differentially expressed with p.adj < .05 or .10. I would like >> the help from folks on this mailing list to introspect the analysis and >> figure out if we overlooked something or data >> >> >> Study Design: >> 4 cases v/s 4 control ( all replicates are biological) >> One sample in the case had very low read counts but I thought that should >> not make a lot of difference in case of DESeq. I did however try to re-run >> the analysis taking that out but no change as far as the #diff exp genes. >> >> Analysis done with : DESeq_1.12.0 >> >> > sizeFactors(cds)TRA00010811 TRA00010812 TRA00010816 TRA00010813 >> TRA00010814 TRA00010817 TRA00010818 >> >> 0.9834455 1.3122005 0.7464998 1.3739400 1.6820479 >> 0.5693563 0.7763204 >> >> >> >> metadata >> >> >> condition libType >> TRA00010811 ablated paired-end >> TRA00010812 ablated paired-end >> TRA00010816 control paired-end >> TRA00010813 ablated paired-end >> TRA00010814 ablated paired-end >> TRA00010817 control paired-end >> TRA00010818 control paired-end >> >> >> >> Attached: >> dispersion plot and MA plot >> >> Please let me know if you need anything more from my end to help me dig a >> bit deeper and understand what might be going on here. >> >> Thanks a lot, >> -Abhi >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor@r-project.org >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor >> > > [[alternative HTML version deleted]]
ADD REPLY •
0
Entering edit mode
hi Abhi, A brief summary of the changes are in the vignette under "4.2 Changes compared to the DESeq package". And there are even more details in the manual pages for ?DESeq and ?estimateDispersions. Mike On Thu, Oct 17, 2013 at 5:32 PM, Abhishek Pratap <abhishek.vit@gmail.com>wrote: > Thanks Lorena and Steve. > > Using the DeSeq2 I was able to get a list of genes after FDR adjustment. We > do some expected functional enrichment there which is great. I would be > curious to know the changes in DESeq2 but I guess the paper is not yet out. > > Cheers! > -Abhi > > > On Tue, Oct 15, 2013 at 1:33 PM, Lorena Pantano <lorena.pantano@gmail.com> >wrote: > > > Hi, > > > > this could happen if your model doesn't explain all your design. There is > > good example in the tutorial about plotting pvalues to see if you get the > > expected distribution. > > > > Sometime I get the same, and in some cases, although FDR is not so low, > if > > I do a heatmap of the top genes, I get my data clustered into the groups > I > > have in my model (cases versus control for instance). So, sometimes there > > is signal even if FDR is higher than 0.1. But you need to be very careful > > and really look your data using other visual representation as others > have > > suggested. > > > > cheers > > > > > > > > > > Lo > > > > > > On Tue, Oct 15, 2013 at 7:50 PM, Abhishek Pratap <abhishek.vit@gmail.com> >wrote: > > > >> Hi Everyone > >> > >> Given the study design( see below) I am somewhat surprised NOT to see > any > >> genes being differentially expressed with p.adj < .05 or .10. I would > like > >> the help from folks on this mailing list to introspect the analysis and > >> figure out if we overlooked something or data > >> > >> > >> Study Design: > >> 4 cases v/s 4 control ( all replicates are biological) > >> One sample in the case had very low read counts but I thought that > should > >> not make a lot of difference in case of DESeq. I did however try to > re-run > >> the analysis taking that out but no change as far as the #diff exp > genes. > >> > >> Analysis done with : DESeq_1.12.0 > >> > >> > sizeFactors(cds)TRA00010811 TRA00010812 TRA00010816 TRA00010813 > >> TRA00010814 TRA00010817 TRA00010818 > >> > >> 0.9834455 1.3122005 0.7464998 1.3739400 1.6820479 > >> 0.5693563 0.7763204 > >> > >> > >> > >> metadata > >> > >> > >> condition libType > >> TRA00010811 ablated paired-end > >> TRA00010812 ablated paired-end > >> TRA00010816 control paired-end > >> TRA00010813 ablated paired-end > >> TRA00010814 ablated paired-end > >> TRA00010817 control paired-end > >> TRA00010818 control paired-end > >> > >> > >> > >> Attached: > >> dispersion plot and MA plot > >> > >> Please let me know if you need anything more from my end to help me dig > a > >> bit deeper and understand what might be going on here. > >> > >> Thanks a lot, > >> -Abhi > >> > >> _______________________________________________ > >> Bioconductor mailing list > >> Bioconductor@r-project.org > >> https://stat.ethz.ch/mailman/listinfo/bioconductor > >> Search the archives: > >> http://news.gmane.org/gmane.science.biology.informatics.conductor > >> > > > > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]
ADD REPLY •

Login before adding your answer.

Similar Posts
Loading Similar Posts
Traffic: 537 users visited in the last hour
Content Search
Users
Tags
Badges
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6