Why should you choose long jumping- distance (LJD) libraries rather than mate-pair libraries, especially for de novo sequencing projects? There is a simple answer to this question: Because the resulting data are much more suitable for de novo assembly and scaffolding.
Why is this so? Mate-pair libraries contain higher percentage of undesired inward-facing read-pairs (if you are not familiar with inward and outward facing reads, just take a look into our FAQs on this subject). These reads are not mate-pair (in other words they are shotgun paired-end). The portion of such reads in the LJD library is greatly reduced.
Furthermore, if using mate pair libraries, it is not clear, if (and if yes, where) there is the changeover within the resulting reads. In other words: A read may contain sequence from one AND the other end without knowing where the changeover is. As a result chimeric reads go into the assembly. This effect is almost completely eliminated if LJD libraries are used, because of the differences in library generation.
Which experiences did you gain with LJD or mate-pair libraries? I’d be happy to hear from you.