Paired-end reads in next-gen sequencing

http://seqanswers.com/forums/showthread.php?t=503
http://www.illumina.com/technology/paired_end_sequencing_assay.ilmn

ECO says ...

For example, you shear up some genomic DNA, and cut a region out at ~500bp. Then you prepare your library, and sequence 35bp from each end of each molecule. Now you have three pieces of information:

--the tag 1 sequence
--the tag 2 sequence
--that they were 500bp ± (some) apart in your genome

This gives you the ability to map to a reference (or denovo for that matter) using that distance information. It helps dramatically to resolve larger structural rearrangements (insertions, deletions, inversions), as well as helping to assemble across repetitive regions.

Structural rearrangements can be deduced when your read pairs map to a reference at a distance that is substantially different from how that library was constructed (~500bp in the above example). Let's say you had two reads that mapped to your reference 1000bp apart...this suggests there has been a deletion between those two sequence reads within your genome. Same thing with an insertion, if your reads mapped 100bp apart on the reference, this suggests that your genome has an insertion.