too high sequencing coverage

[Antony03]

Hi,

I sequenced by Illumina MiSeq some phage genomes. Normally, with bacterial genomes, I have an average sequencing coverage of 75-90x. But with the phages (much more small genomes) the coverages are more than 1500x.

I have read at some places that a too high coverage might be "problematic". Is someone can explain to me why?

Should I perform a random subsampling of my reads to assemble at a lowest average coverage (~100x)?

Thank you,

Antony

[Brian Bushnell]

It depends on what you want to do. Most assemblers have heuristics geared toward much lower coverage, on the order of 100 or less. Super-high coverage may give very fragmented assemblies; with low coverage a given error is typically unique, but with high enough coverage the same error may be seen multiple times, and actually assembled. That will create false branches on a DeBruijn graph that may be cut, causing fragmentation. Some assemblers may also simply not assemble very-high coverage areas for other reasons.

If you have fairly uniform coverage, subsampling is a good option. If you have variable coverage, normalization is a better choice.