What do you do with unused reads after de novo assembly?

[morning latte]

Dear all,

I have metagenomic data which was produced from Illumina HiSeq 100 PE sequencing. After de novo assembly using velvet with a range of kmers, I came up with one kmer. When I looked at the percent of reads used for assembly, 70% of reads were actually used. I got curious what people are doing with the rest of 30% of unused reads and looked into threads related to this but haven't got a clear idea. Could anyone give me some advice or direct me any publication? Thanks!

[Brian Bushnell]

Regardless of which reads were used, you might try mapping all input reads to the assembly, then assembling the pile of reads that did not map with some score cutoff. It's difficult for assemblers to handle metagenomes with exponentially-distributed coverage and this makes it a lot easier to assemble the high-coverage organisms in one pass and the low-coverage organisms (if possible) in a second pass. Iterate as needed, as long as coverage is sufficient for assembly.