Help me!!!!! low % of properly paired reads!!!

[Trudy]

Hi all,
I have an important question:
I have 75x2 reads from Illumina HiScan and I preformed:

-alignment with TopHat
-samtools flagstat to know the % of properly paired reads

This % very low, around 1% for all samples........why???

I have also increased the -r option (mate inner dist), from 85 to 1000 but there're no significant differences!

My only ipothesis of these situation is a very low quality of reverse strand because during the run we had problem with enzyme and the boxplots of reverse reads had low quality associated value. Is it possible??? Low quality of reverse strand reads can influence the % of properly paired reads??

I think also that downstream analysis are compromised......Help me please if sameone knows this situation!

[simonandrews]

Assuming you're aligning to genomic sequence then any pairs which fall into different exons (which will probably be quite a lot of them), will have effective inner distances much bigger than 1kb due to the added size of the introns you'll be spanning.

Looking at mate pair distances on the genome doesn't make much sense for RNA-Seq data. If your improper pairs come from the pairs mapping to the same strand, then that's cause for concern, but as long as your pairs fall into the same gene then for RNA-Seq that's probably enough.