Experimental design

[Arvind Bhagwat]

We want to investigate if certain treatment (nutrition supplement) changes mouse fecal microbiome, and if it does then how long does it take to change the biome? We have sample from -3 days to 21 days (every alternate day collection). We will be doing 16 S rRNA gene seq/MiSeq. What I am not sure is how best to use 96 indexes? That is to say whether to:
1. take 3 mice (individually indexed) per day per treatment and do more samples over 21 days or
2. take 5 mice (individually indexed) per day per treatment and do samples but from intermittent days.
Your comments / critique welcome!

[dariober]

By gut feeling I would favour option 2 (more mice but fewer time points, right?) but obviously it depends on how much variation you expect between mice and between time points. If mice are very very similar (which instinctively I doubt) than there is little point in doing 5 instead of 3. Similarly, how important is it to have closely spaced time points? Do you expect sudden changes from one day to another?

Maybe see if the RNASeqPower Bioconductor package is any useful

[thermophile]

Why limit yourself to just 96 indices? If you're trying to do this in just one run, do more mice and every day collection. 5 individuals is pretty tiny sample size, I think you're going to find that 5 mice are changing but not in a uniform manner. Schloss lab apparently multiplexed 384 (including controls) per MiSeq run, I haven't had much luck in uniform distribution of indices when I go above 150 but I'm trying to up it. You will likely loose a few samples in the sequencing run if you go very high but I think that more mice will give you much more power even if you only have duplicates for some mice for some days.