So I have 454 sequences from several samples which I would like to compare. No separate 16S data was generated. What would be the best way for OTU diversity analysis? I do have MG-RAST results for these data, including GreenGenes/RDP/Silva. My concern is that the number of OTUs identified this way is rather small. And yes, I did denoise the data.
If I understand your question correctly, you have shotgun metagenomic 454 data. If so, three options spring to mind: MEGAN, CARMA and the Galaxy Metagenomics Workflow. I've had good luck running and modifying the latter, but don't have experience with the other systems. If these are on target, it would be good to know which works best for you.