Hi all. I've been working with comparing expression values among different isoforms of loci in Drosophila over a number of samples and have noticed that there's a consistent bias in RPKM values wherein counts are overwhelmingly allocated to longer isoforms (I've seen this with both Cufflinks and RSEM). If anyone else has encountered this, I was just looking for a quick opinion: Would it be legitimate to subtract the mean or median expression of an isoform over my samples and compare residuals among isoforms to get rid of the bias?
Alternatively, if anyone has seen this discussed in the literature, I'd love to read about it. My concern is that such biases are completely interfering with proper isoform quantification in an unpredictable way.
Thanks,
Carlo
Alternatively, if anyone has seen this discussed in the literature, I'd love to read about it. My concern is that such biases are completely interfering with proper isoform quantification in an unpredictable way.
Thanks,
Carlo