beta diversity depth

[tsangkl]

Hi all,

I am using QIIME to analyse my 16S metagenomic sample.
I am going to compare the bacterial community of 2 group of sample and each group composed of 10 sets of samples

I have merged, trimmed and filtered my sequences and the no. of read for my 10 samples are:

Group 1: 88322, 131727, 150013, 169207, 177499, 193288, 197006, 200491, 201732, 229860
Group 2: 115444, 127776, 172511, 172573, 179295, 181659, 186582, 200619, 201387, 212047

I have subsample the reads to multiple rarefaction (sample size from 1000 to 88000 with step size of 2000) to calculate the alpha diversity with parallel_multiple_rarefactions.

I am going to use the jackknifed_beta_diversity to see if there is difference between 2 group of samples. Unlike alpha diversity, it seems that only a single depth is allowed to compute the beta diversity. I would like to ask what is the best sequence depth for compute the beta diversity and plot the principal coordinates analysis for these sets of data ? Between, is it a good idea to remove the data with only 88322 reads which is relative fewer reads?

Thanks for answering the long question.

[fanli]

88k reads is generally more than sufficient to saturate each sample. Look at your alpha diversity rarefaction plots - do they plateau way before 88k? If so, then you probably have a good representation of your population at lower read counts.

[tsangkl]

I have attached the rarefaction plot, the increase in observed OTU numbers slow down when more sequences, but seems not reaching a plateu??

Is the observed species over-estimate? I see the observed OTU from other papers usually below 1k.

[fanli]

Yeah, those do seem high but it depends on your sample (e.g. bacteria-rich soil). How are you doing the OTU picking? Are you filtering the OTU table afterwards, for example to remove really low abundance species? We typically remove anything at < 0.005% abundance and this leaves us with a few hundred OTUs.

You want to run the rarefaction plot out to >88k to see if you want to remove the sample with fewest reads, right?

[tsangkl]

I pick the otu by pick_open_reference_otus against greengenes 97% clustering 16S reference set (remove singleton by default):

pick_open_reference_otus.py -i input.fasta -o otus -r 97_otus.fasta

And I have taken your advise by removing < 0.005% abundance reads and still got over 4000 OTUs.

Yeah, those do seem high but it depends on your sample (e.g. bacteria-rich soil). How are you doing the OTU picking? Are you filtering the OTU table afterwards, for example to remove really low abundance species? We typically remove anything at < 0.005% abundance and this leaves us with a few hundred OTUs.

You want to run the rarefaction plot out to >88k to see if you want to remove the sample with fewest reads, right?

[GA-J]

Hello, Fanli, I like your result of 16s V4 Miseq run. I would like to know how much you loaded? And the size of your library is ?

Thanks.