Tweet
I have a rather basic question regarding EDASeq that I have not found the answer to in reading the vignettes.
Its pretty clear, how to work when one has technical replicates in individual lanes of the same library. However I have two conditions, each with three biological replicates, multiplexed and sequenced across three lanes.
Where I get hung up then on is the best way to approach this. Should I be combining counts from all three lanes for each bio rep and normalizing that? Should I separate reads for each lane and normalize for both samples and lanes? In this latter case, how do I then recombine my technical replicates for differential expression?
I have used EDASeq in the former, where each sample is the combined results of all three lanes and each sample treated as if it were its own lane. This has a huge impact on differential expression compared to previous analysis using only DESeq and EdgeR without correction for GC bias, with the result that it actually matches qPCR data better, but I am worried that this is still an inappropriate setup that is potentially giving me wrong results.
Its pretty clear, how to work when one has technical replicates in individual lanes of the same library. However I have two conditions, each with three biological replicates, multiplexed and sequenced across three lanes.
Where I get hung up then on is the best way to approach this. Should I be combining counts from all three lanes for each bio rep and normalizing that? Should I separate reads for each lane and normalize for both samples and lanes? In this latter case, how do I then recombine my technical replicates for differential expression?
I have used EDASeq in the former, where each sample is the combined results of all three lanes and each sample treated as if it were its own lane. This has a huge impact on differential expression compared to previous analysis using only DESeq and EdgeR without correction for GC bias, with the result that it actually matches qPCR data better, but I am worried that this is still an inappropriate setup that is potentially giving me wrong results.