I would really appreciate someone's input on some issues I am having with my cuffdiff output for RNAseq data and how it compares with genes I have viewed on IGV (or a similar browser) -
1. when I look at my differentially expressed transcripts file (generated using ensembl hg19 as a reference with chr added on to obtain results with ensembl gene names) and search for specific genes that I am interested in I can not find them in my cuffdiff output file - even though I can visualize these genes on IGV and they look obviously differentially regulated. Also, given that the cuffdifff output for differentially expressed transcripts does list all trascripts, including the ones that have not significantly changed, wouldn't transcripts for these genes be listed anyway, even if my visual ballparking on differential regulation is not statistically significant? I would really like to know why I am missing genes from my cuffdiff output.
2. do you all get a good correlation between the top differentially expressed transcripts/genes generated from cuffdiff and how the data looks when visualized on IGV - ie. do your upregulated transcripts really look upregulated when visualizing? I found that while some validate visually, some do not which is confusing....
3. when visualizing on a browser, and if different transcripts for one gene are regulated differently - ie. some are up in your treated sample but some are done for the same gene - how can you tell which transcriptID from cuffdiff corresponds with what you are seeing?
Thank you all for your help!
Kurinji
Graduate Student Johns Hopkins School of Medicine
Kurinji,
- when I look at my differentially expressed transcripts file (generated using ensembl hg19 as a reference with chr added on to obtain results with ensembl gene names) and search for specific genes that I am interested in I can not find them in my cuffdiff output file - even though I can visualize these genes on IGV and they look obviously differentially regulated. Also, given that the cuffdifff output for differentially expressed transcripts does list all trascripts, including the ones that have not significantly changed, wouldn't transcripts for these genes be listed anyway, even if my visual ballparking on differential regulation is not statistically significant? I would really like to know why I am missing genes from my cuffdiff output.
It's not possible to answer this question in general because it's specific to your analysis; in particular, your use of a reference annotation file is going to influence Cuffdiff's outputs. You might try using positional information rather than gene names when searching through Cuffdiff files as the gene short name/ID is only used for known transcripts/genes.
More detailed questions are probably best directed to the Cufflinks authors: tophat.cufflinks@gmail.com
- do you all get a good correlation between the top differentially expressed transcripts/genes generated from cuffdiff and how the data looks when visualized on IGV - ie. do your upregulated transcripts really look upregulated when visualizing? I found that while some validate visually, some do not which is confusing....
Cufflinks uses multiple statistical techniques to estimate FPKM and differential expression; in some cases, it may not be possible to visually observe differential expression amongst transcripts. Alternatively, setting additional parameters (e.g. normalization) may lead to results that match what you're looking for (visually or otherwise).
- when visualizing on a browser, and if different transcripts for one gene are regulated differently - ie. some are up in your treated sample but some are done for the same gene - how can you tell which transcriptID from cuffdiff corresponds with what you are seeing?
This information can be found in Cuffdiff's transcript FPKM tracking and differential expression testing files.
Good luck, J.
galaxy-user@lists.galaxyproject.org
-
Jeremy Goecks -
Kurinji Pandiyan