I'm working with a transcriptome assembled from RNA-Seq data on a time course experiment in infected and uninfected cells, three replicates per condition/time. Initially I had calculated DE on the whole transcriptome, I then filtered out transcripts based on BLAST results and having a maximum FPKM (for any condition/time) of > 0.5.
However, when I went back and looked at the DE results, I noticed that this filtering was removing up to 90% of the DE transcripts. That is, only 10% of transcripts with sufficient FC/adjusted pval had sufficient BLAST/FPKM.
I'm currently trying the opposite approach, where I first filter transcripts via BLAST and then calculate DE on just these transcripts using all of the read data. However, in the process of doing this I noticed bowtie2 overall alignment rates were 40-50%, very low. That makes me think that this approach is incorrect.
Any suggestions?
