Hello everyone,
We are working with some data in lab in which we are comparing two different treatments but did not have the sample means necessary for replicates. When I run cuffdiff I only get 30 or so transcripts that are significantly DE out of the few thousand or so. The weird part is that all of these transcripts it finds significant are expressed in one sample at a fairly high level and not expressed in the other sample at all. When running DESeq I am seeing no significance in any transcripts for DE analysis. This information seems rather non-essential to be kind and I know our problem is no replicates. Telling me that a transcript that isn't even expressed in one population but highly in the other(cuffdiff) is differentially expressed is stating the very obvious.
My thoughts on how to approach our data are as follows and I was looking for some more RNA-Seq savvy people to let me know if this is acceptable or not as we are fairly new to RNA-Seq and this is our first sample to sample comparison.
What I proposed we do with our data is to take cufflinks FPKM values and just simple evaluate ratios, calling anything with say 2:1 ratio DE and having higher ratios (ex. 5:1; 10:1) in separate categories. We have looked at the data in this manner and tried to validate with our little bit of leftover sample on qPCR to determine if these ratios are indeed truly represented in the sample via validation.
Could this be an acceptable alternative to cuffdiff/deseq/etc for not having replicates and those analyses offering little to no useful information? Is validation via qPCR essential or could these ratios be reported as is?
I feel our data is still very useful exploratory data as RNA-Seq has never been done to compare these 2 sample types but the DE analysis has been lacking due to our lack or replicates.
Thank you in advance for any input,
-C
We are working with some data in lab in which we are comparing two different treatments but did not have the sample means necessary for replicates. When I run cuffdiff I only get 30 or so transcripts that are significantly DE out of the few thousand or so. The weird part is that all of these transcripts it finds significant are expressed in one sample at a fairly high level and not expressed in the other sample at all. When running DESeq I am seeing no significance in any transcripts for DE analysis. This information seems rather non-essential to be kind and I know our problem is no replicates. Telling me that a transcript that isn't even expressed in one population but highly in the other(cuffdiff) is differentially expressed is stating the very obvious.
My thoughts on how to approach our data are as follows and I was looking for some more RNA-Seq savvy people to let me know if this is acceptable or not as we are fairly new to RNA-Seq and this is our first sample to sample comparison.
What I proposed we do with our data is to take cufflinks FPKM values and just simple evaluate ratios, calling anything with say 2:1 ratio DE and having higher ratios (ex. 5:1; 10:1) in separate categories. We have looked at the data in this manner and tried to validate with our little bit of leftover sample on qPCR to determine if these ratios are indeed truly represented in the sample via validation.
Could this be an acceptable alternative to cuffdiff/deseq/etc for not having replicates and those analyses offering little to no useful information? Is validation via qPCR essential or could these ratios be reported as is?
I feel our data is still very useful exploratory data as RNA-Seq has never been done to compare these 2 sample types but the DE analysis has been lacking due to our lack or replicates.
Thank you in advance for any input,
-C
Comment