Hi Rohan, quoting from illumina's 16S library prep guide "The MiSeq run output is approximately > 20 million reads and, assuming 96 indexed samples, can generate > 100,000 reads per sample, commonly recognized as sufficient for metagenomic surveys". Not sure how this applies with lower numbers of samples?

Hi Rohan, yes this study is looking at the microbial community in fish GI tract, by analyzing 16S amplicons with the MiSeq. The read counts are ok up to OTU assignment, it's when I apply 97% identity filters that they drop. I still get an excellent idea of the diversity of these communities with those reads. I'll see what others think

We have been able to generate >120K sequences per sample.

Which means there has been some issue probably with amplicon pooling prior to sequencing or have addded more than 60% phix.

There is a detailed protocol including wet lab SOP in AEM from P.Schloss, this is actually more useful and realistic than the Illumina protocol.

There are also some issues specifically for 16S on illumina that one needs to be

aware of.

Cheers

Dear Jay

Thank you for your input. Just to clarify, after applying 97% identity filters, do you get the 120,000 reads? I am using v4 primers.

Cheers

hi Philip,

We had multiplexed about 25 samples using the V4 region and managed to get 120K plus reads per sample for a paired end read.

We did get low sequences once but it was related to the phix spike and pooling issue that resulted in poor quality reads that did not pass QC.

Thanks Jay, 120,000 before or after 97% identity filter?

Cheers

We had > 120K sequences post qc steps and chimera and non 16S sequence removal.

As outlined in the miseq SOP analysis steps here,.

http://www.mothur.org/wiki/MiSeq_SOP

Hi Rohan

I still believe that the MiSeq is the better instrument for 16S metagenomics. Maybe i'm wrong, but the data I have are nice, giving me a good idea of the microbes in my samples, and it seemed to work for Jay too.

Hi Rohan,

it depends what you are looking for, if you know what ur looking for ie have a candidate,,,then you can do targeted sequencing and for that 454 should be enough. one also has to calculate the depth of sequencing, expected diversity and multiplexing strategy.

But if your project is exploratory, a hiseq might be the best way forward, as long as you have the computational resources to crunch through the data and do the necessary correlations.

so some studies are

https://www.cell.com/trends/molecular-medicine/abstract/S1471-4914%2813%2900148-2

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0039743

Just make sure you run a test sample to check out the inherent noise in ur hiseq for 16S and dont change the machine/chemistry or hard code matrix if you are using one, until your project is done.