reps and analysis

[Vale16S]

Hi all!

I am doing a 16S amplicon analysis on soil samples with 5 different treatments.
I have 9 replicates per sample.
I got from illumina base space a file containing all the reads for all different taxonomic levels. Right now I am working on the class one.
I normalized the reads to the total number of reads. now I would like to verify the variability within my treatment between the reps, and obtain a single data set to work with per traetment.

Could you suggest me how is the best way to proceed with the analysis testing the variability, for istance, and then combine the sequences coming from the replicates into a unique sample sequence data set?
Could you suggest me the best method to analyze the variability between my replicates?

Do you have ideas?

Thank you very much!

[fanli]

Generally you don't want to discard replicate information. All/most of the downstream analyses that you want to do can and should be done with all of your replicates intact.

See the QIIME tutorial here:

http://nbviewer.ipython.org/github/biocore/qiime/blob/1.8.0/examples/ipynb/illumina_overview_tutorial.ipynb

A PCoA plot stratified by your five different treatments might be a quick and easy way to convince yourself of the reproducibility of your replicates.

[Vale16S]

Hi,
Thank you fanli for replying.
I have a couple of doubts still about reps treatment for 16S sequnecing.
I didn't mean to discard the replicates data. I would like just to ask how to deal with them.
I see the PCoA plot as a tool to evaluate the reproducibility of the reps. What will you do then?
would you combine the replicates reads ? would you sum the replicates reads? how will you proceed in order to obtain treatments sets?

Thank you a lot!

Cheers

[fanli]

I suppose you could combine the replicate reads if you wanted a single list of OTUs for each treatment. What exactly are you trying to compare with your data? If you want to compare your treatments, it makes more sense to compute diversity/population statistics for each replicate and then compare those between treatments (e.g. a 9-vs-9 comparison).

[Vale16S]

HI FAnli again,
I would like to compare the treatments, taking into account the intrinsic variability within the samples (due to DNA extraction or due to field).
I thought to have a single data set per treatment. maybe it is not the right way.
Which tool will you use for the 9vs9 comparison?
thnks

[fanli]

Hi Vale16S,

I've been using QIIME to compare alpha- and beta-diversity statistics between treatment groups. See http://qiime.org/scripts/compare_alpha_diversity.html.

From the man page:

For example, if your mapping file had a category called ‘Treatment’ that separated your samples into three groups (Treatment=’Control’, Treatment=’Drug’, Treatment=‘2xDose’), passing ‘Treatment’ to this script would cause it to compare (Control,Drug), (Control,2xDose), (2xDose, Drug) alpha diversity values. By default the two-sample t-test will be nonparametric (i.e. using Monte Carlo permutations to calculate the p-value), though the user has the option to make the test a parametric t-test

So you'd be doing a 9vs9 t-test, essentially.