How to interpret "insert size" after RNAseq

[raphael123]

Hi,

I am really new in NGS but I have looked for RNAseq data, aligned with tophat. I used then picard to look at different statistics on this dataset and specifficaly what is called "insert size". It should be define as the distance between two potential mate pair.
First : How mate pair are linked ?
Second : What is usually do with this data ?

But my main question regarding this histogram :

http://nasty.mushroom.free.fr/trucs/insertsizehistogram4forum.pdf

When an insert size is big, it could mean the gene in dna contain big intronic region. Where is it discuss ? What do you think about that ?

I would enjoy any comment on this !

[dpryan]

Well, if the observed insert size is greater than that used to create the library then the fragment likely contains a splice site. I've never seen a need to look at this with RNAseq data, at least when everything is mapped with genomic coordinates (I suppose that looking at insert sizes when things are in transcriptome coordinates might be useful for QC).

[raphael123]

It could be interesting :

1 to choose a threshold (like 50kB) in order to filter for false read mates.
2 to have a reference of genes to check if the 2 read mates are in the same one.