reads mapping to overlapping genes?

[beliefbio]

Many reads are mapped to positions where overlapping genes reside. So we overestimated the number of reads mapping to some genes. How to deal with this problem?
Thank you for your time!

[simonandrews]

Originally Posted by beliefbio

Many reads are mapped to positions where overlapping genes reside. So we overestimated the number of reads mapping to some genes. How to deal with this problem?
Thank you for your time!

It depends on the structure of the genes, but we've started calculating our counts as bp of read per kbp of transcript, but calculating this value separately for every exon, and then taking the mean or median value of all exons as the overall value for the gene. This seems to cut down on problems which only affect a small part of the gene.

[steven]

Originally Posted by simonandrews

It depends on the structure of the genes, but we've started calculating our counts as bp of read per kbp of transcript, but calculating this value separately for every exon, and then taking the mean or median value of all exons as the overall value for the gene. This seems to cut down on problems which only affect a small part of the gene.

Yes, this should also reduce the impact of alternative splicing. One question: if different transcripts use the same exon, do you count the corresponding value several times when computing the mean or median, or just once?

[eszter.ari]

Take a look at our read counter tool that does not count read-pairs mapped to the same-strand overlapping part of genes: http://seqanswers.com/forums/showthread.php?p=134850
https://code.google.com/p/recog/