comparative metagenomics

[dnusol]

hi all,

We are designing a comparative metagenomics experiment where we may run 3 different conditions with several individuals per condition. I am trying to see what would be an appropriate way for analysis and I got down to two ideas.

On one hand I could pool all reads from the different individuals per condition, to get a "consensus" metagenome per condition. This way, I reduce the analysis down to comparing three metagenomes. I think that pooling the replicates will help me detect low represented organisms, but on the other hand I lose biological variability. This analysis would probably go down the qualitative way.

On the other hand, I can assemble one metagenome per individual and then compare them between groups taking into account the reps. This analysis would then go down a (semi)quantitative way. But I am not sure how to handle that, what would be the data to compare or how to get the data at all.

I have seen that MEGAN 2.0 seems to do a nice job on comparing metagenomes, but it seems to be a one vs one approach, more like the first option of analysis.

See for example:

http://www.biomedcentral.com/1471-2105/10/S1/S12

It would be great if anyone could share ideas, software, references or experience on such an experiment.

Cheers

Dave

[dsher]

Hi
It seems to me that it is always better to treat each replicate separately, at least at the stage of library preparation and sequencing. Not being able to describe the biological variability is a real setback in many omics studies, maybe even at the level fo bad experimental design. You can always combine the libraries in the analysis. It does cost more, though.

BTW, this paper has some interesting ideas on how to look at biological replicates: http://www.nature.com/nature/journal...ture13786.html

Good luck
Daniel