I'm having troubles understanding something which I though I knew well (hard times, these). I always knew that, while Cufflinks returns FPKM, expected counts and other things, HTSeq returns raw counts. This should mean that in the output of HTSeq I should have the counts of the fragments that were mapping on my particular gene. Right?
I am graphically and command-linely assessing the fact that I have indeed around 30 fragments mapping on a gene in the wild type condition and 0 in the treatment, but htseq returns only 1 in the wild type and 0 in the treatment. This gene is knocked-down, so I know it's down-regulated also from wet lab experiments.
EDIT: Three questions:
How does HTSeq handle reads that exceed borders of the considered regions but overlap them? My guess is that it doesn't count them.
I always thought that I understood correctly this: the count should just be a count of the fragments (in case of PE reads), without any normalization, right?
What the hell is going on?
The command is:
python /usr/bin/htseq-count -f bam -r pos -t exon -i "ID" alignment.bam gene-set.gff3 > stdout 2> stderr
Thank you in advance for the help!