Megan for contig or gene classification

[pschwien]

Hi SEQanswers experts,

I've used megan extensively to taxonomically classify reads from metagenomes which is, according to the initial publication, one of the main reasons the software exists. However, I wonder if it also suitable to classify genes from assembled metagenome contigs - or even whole contigs. I am only interested in a very high level (Domain) classification to estimate the contribution of archaea, bacteria, eukaryotes, and virusses/phages to the metagenome. Since megan only uses the blast result files there shouldn't be a technical problem but I am worried about the theoretical concepts behind the algorithms (e.g. LCA), will megan still yield unbiased & reliable results?

If so, I would appreciate advice for which blast[n|p] and megan parameters in order to yield best results for gene or contig blasts.

Lastly if megan turns out to be not applicable to my problem, do you know about software that would be better suited?

Thanks in advance

[tsa]

Sequence assembly is a non-linear step, thus your results would be heavily biased.

[rhinoceros]

You could submit your sequences to some online service like MG-RAST or IMG/M. Or you could sort your blast result for best hits and extract the domain level classification straight from that. For example -outfmt '6 std sskingdoms" (I recall) and then grep from the last field counts of A/B/E/V (you would also need the taxdb for this). IMO 1e-6 is a good evalue for blastp against nr. I usually also apply the -use_sw_tback, -seg yes and -soft_masking true flags because of some paper I read which recommend these but I don't really know how applicable they are to the latest version of blast..