MG-RAST settings, thoughts and suggestions

[JJenks]

Hi All,
I'm in the process of using MG-RAST (among other methods) to determine the taxonomic origin and function of reads from a shotgun study using Illumina reads. I've carried out a few different analyses using the online workbench, and just wanted to hear the opinions of others who have used it. I'm trying to work out if it is more sensible to have a fairly lax e-value cutoff (i.e. the default) and increase the % match and alignment length, or to up the e-value.
I'm also still slightly confused about the definition of the e-value within MG-RAST too. I'm assuming that it's the same as BLAST, i.e. taking into account sequence length, quality of alignment and database size the e-value is representative of odds of a match by chance. Can anyone confirm this?
Sorry if this seems confused, I'm trying to get my head around all of the concepts to ensure the science is robust.

Thanks in advance!

[JJenks]

Just a quick update, I've since found some further details on MG-RAST, and can confirm e-values are identical to the NCBI definition found here: http://tinyurl.com/cjrjdcy. I'm finding alignments of 30-35 bp and at 95%+ identity, with an e-value of 1e-10. These would appear to be fairly good matches to me, although perhaps the alignment length is a little short.

Thanks again.