insert size

[adrian]

Dear group:
I need some help in understanding insert size concept.

I have a targeted exome sequencing data using paired-end approach with 76 bp. I have lots of duplicates in the sam file. I should use rmdup with insert size correctly mentioned.

i was told by technician that insert size is between 150-300 bp.

When I see 9th tag which is inferred insert size in sam file, I have lots of numbers that range from 0 to 100,000.

Since the experiment is done with an insert size 150-300 bp, and BWA inferred insert size has lots of ranges, what number should I use in using rmdup. Heng Li recommends that we should use correct insert size always. If I have range from 150-300 (technician) and SAM file inferred sizes are spanning across wide ranges, Which insert size should I select to remove duplicates and call SNPs.

OR should I make sets of reads that fall into certain ranges and call SNPs in each bin.

Also what is inferred insert size '0' mean and what is 345,039 mean.

thanks
Adrian

[krobison]

Many if not all of the very large insert sizes are probably due to falsely aligned reads to repeats (LINEs, SINEs, Satellite, etc); take a few of them and look to see whether both reads actually align to non-repeat DNA.

Also, take a sample from your SAM file (ideally random) & look at the size distribution -- you may see a long tail of weird sizes, but you'll probably see most of the counts in a distribution around what the technician said. It's a worthwhile check on the library in any case & easy to generate the data table with a little bit of perl (or even UNIX shell commands).