Insert size

[billstevens]

Hi guys,

What do you guys think about library size? I'm doing 100 bp PE reads, and I just did a run with an insert size of 200 bp. I got very good mapping 90%, but now i'm wondering if it was so good because there was overlap? I talked to the Illumina rep who mentioned that the power of it is to have a read, a gap, and then a read, so should I have selected 300 bps? This seemed more art than science, but what should I be looking for in making this determination?

[pbluescript]

What is your goal for doing PE sequencing?

[billstevens]

Mostly differential gene expression, differential promoter use. I'm not really looking for novel transcripts or that stuff. Its human epithelial cells by the way.

[pmiguel]

Probably makes little difference in your case.

The advantage of maximizing spacing between read pairs by creating larger insert size library molecules is that if one read end happens to map in a repetitive block, the other has a greater chance of mapping outside that block (in single copy sequence) if if maps farther from the first. That is, repetitive blocks have some maximum size, if the distance between read pairs exceeds this distance you have a very good chance of being able to anchor your read pair. In principle, only one read is necessary to anchor the pair.

That is "in principle". Depends on your mapping engine whether, in practice, you will actually see any benefit. Also, transcriptomes will generally have fewer "repetitive" blocks than, say, genomes.

There is a down-side to larger inserts -- shorter inserts amplify more efficiently and thus give more robust signal.

--
Phillip