Are multiple insert sized PE libraries worth the cost

[Potjie]

We are planning to sequence a large number (>10) of genomes. All are approximately 40Mb in size.

We are planning to sequence them on the Illumina Hiseq platform.

Obviously we would like to get the best assemblies possible from the data, however we do not have the funds for Mate-Pair sequencing for all of the genomes.

We are considering sequencing 2 Pair-end libraries of different insert sizes (e.g. 300kb and 600kb) for these genomes.

Has anyone done this before? And would you say that the cost of doing a second PE library was worth the improvement to the genome assembly?

[Brian Bushnell]

If you can't afford LMP libraries, you may still be able to improve your scaffolding with the longest possible fragment library, around 800bp. Whether multiple libraries are helpful depends on the assembler. For example, AllPathsLG tries to join paired reads, and to do so, it's best to have a shorter insert, preferably overlapping. So (for 2x150bp reads) a 270bp-insert library and 800bp-insert library might be helpful, though I don't have any hard data on how good 800bp is for scaffolding (it's still pretty short).

But unless you or the assembler are going to try to join the reads, I doubt a second shorter insert library would be much help, so I would tend to pick a single 800bp library and save the extra money (or spend it on LMP libraries for the genomes that end up in the worst shape). Though again, I don't have any hard numbers on this.

You could, of course, generate some synthetic data from a reference genome and see how a second library with a different insert size improves the assembly.

[Potjie]

Thanks for the info! Its much appreciated.