structural variation and read length

[upenn_ngs]

What is the current thinking about paired read length (100bp or 150bp) and detection of structural variants? Are the longer 150bp reads advantageous? And how might the library insert size influence detection?

[krobison]

Once you've successfully mapped each end, longer reads won't help, except you increase your odds of actually reading across the breakpoint. Of course, once the two reads are meeting in the middle, you aren't helping in that department either.

Library insert size will have an effect, with longer obviously better -- for a given library you can calculate the depth of coverage of the genome as a whole, and then estimate your probability of seeing a breakpoint given the amount of coverage. But if longer inserts mean lower numbers of reads, then the cost-benefit curve could be modeled.

For looking for the whole spectrum of structural variants, then your ability to detect is affected by (a) distance between your reads relative to the original genome (b) your accuracy in estimating that distance (c) the size of the variant.

[upenn_ngs]

thanks krobison, but i'm a bit dense. could you spell out your explanations with more detail. personally i am interested in detecting structural variation from exome data.