16s rRNA gene sequencing depth and reference database

[illuminaGA]

Dear Guys

I'm starting a metagenomic analysis on the rats fecal microbial community.
But our lab doesn't have any experience about this work.
We will have 96 rats fecal samples and plan to run 16s rRNA gene sequencing by Illumina Miseq.

I still confused about some question after reading papers and protocols.

1.About the depth. How much data is enough for per sample? Some one only got 3M/per sample.

2.How to choose the right reference database for data analysis?

Any help would be greatly appreciated!

[TiborNagy]

2. The right reference database should be regularly updated and well curated, like Silva.

[bstamps]

1. How much depth is sample dependent. Do you have a hypothesis as to the diversity of the environment? Do you have a question/hypothesis that can be tested with more/less depth? 3 million reads per sample is approaching stupid-high levels of depth for a 16S survey in most environments.

2. The "right" database is open for debate. Our lab uses Silva because of its size and the large number of uncultivated candidate groups that are well represented in it (We work with environmental samples). Many people use the GreenGenes DB, either because it's in the Knight/Caparaso SOP to use it (And thus your data fits nicely into their pipe) and/or because clinical samples can be described very well with said DB.

FYSA you have a 16S (Not little s) amplicon survey library, not a metagenome... just a sticking point with me.

[illuminaGA]

Thank you so much!

[illuminaGA]

1. How much depth is sample dependent. Do you have a hypothesis as to the diversity of the environment? Do you have a question/hypothesis that can be tested with more/less depth? 3 million reads per sample is approaching stupid-high levels of depth for a 16S survey in most environments.

2. The "right" database is open for debate. Our lab uses Silva because of its size and the large number of uncultivated candidate groups that are well represented in it (We work with environmental samples). Many people use the GreenGenes DB, either because it's in the Knight/Caparaso SOP to use it (And thus your data fits nicely into their pipe) and/or because clinical samples can be described very well with said DB.

FYSA you have a 16S (Not little s) amplicon survey library, not a metagenome... just a sticking point with me.

Hi bstamps

Thank you so much for your information.

We want to identify the gut microbiome of our transgenic rats and control rats ,then compare the difference. We plan to sequence the V3 and V4 region. Do you think the 16s rRNA sequencing can achieve our goal?

If the 3 million reads is "stupid-high levels" .What is the reasonable depth for bacterium identification?

[bstamps]

Well, once again it depends. Sure, V3/V4 sequencing will allow you to ID the microbial community to the level of family (Maybe genus), which given enough samples and experimental variables would allow you to do plenty. We routinely get 40,000 to 100,000 reads per sample and consider that to be enough depth in sequencing for 16S amplicon on the Illumina platform. If you're looking for some ultra rare taxon (which is a hazy path to go down with small fragment 16S sequencing), then yeah, you might want much more depth, but if the question is "Do we observe a shift between my control and my experimental samples" getting on average 50,000 reads per sample should get you to your finding no problem.

I guess the short answer is: Use short reads to ID mid to major shifts in the microbiota of a dataset, which is what it sounds like you want to do. And at 50,000 , 100,000 , or 200,000 reads per sample you should be doing great.

[Genohub]

V4 library prep and sequencing

I agree, 50,000-100,000 reads is likely enough for your project. We've had good experience with the Green Genes db: http://greengenes.lbl.gov/cgi-bin/JD...al/nph-16S.cgi

If you're looking for providers who can make 96 V4 or other variable domain libraries and sequence them at 50,000 reads/sample: https://genohub.com/shop-by-next-gen...32a9b406a7ed4b

- Genohub

[illuminaGA]

Well, once again it depends. Sure, V3/V4 sequencing will allow you to ID the microbial community to the level of family (Maybe genus), which given enough samples and experimental variables would allow you to do plenty. We routinely get 40,000 to 100,000 reads per sample and consider that to be enough depth in sequencing for 16S amplicon on the Illumina platform. If you're looking for some ultra rare taxon (which is a hazy path to go down with small fragment 16S sequencing), then yeah, you might want much more depth, but if the question is "Do we observe a shift between my control and my experimental samples" getting on average 50,000 reads per sample should get you to your finding no problem.

I guess the short answer is: Use short reads to ID mid to major shifts in the microbiota of a dataset, which is what it sounds like you want to do. And at 50,000 , 100,000 , or 200,000 reads per sample you should be doing great.

Thank you so much!

If we sequence V3 or V4 region only, is this enough for beactria identification?

[illuminaGA]

I agree, 50,000-100,000 reads is likely enough for your project. We've had good experience with the Green Genes db: http://greengenes.lbl.gov/cgi-bin/JD...al/nph-16S.cgi

If you're looking for providers who can make 96 V4 or other variable domain libraries and sequence them at 50,000 reads/sample: https://genohub.com/shop-by-next-gen...32a9b406a7ed4b

- Genohub

Thank you for your info.

[trexbob]

Thank you so much!

If we sequence V3 or V4 region only, is this enough for beactria identification?

This depends on the genus/species and how diverse they are in the region that is being looked at. Data analysis parameters figure in as well. The higher % ID you pick OTUs at, the more species will be differentiated, for good or bad. example: two closely related species may be grouped together if you pick OTUs at 97%, where they may be classified as different OTUs at 99%.

[illuminaGA]

This depends on the genus/species and how diverse they are in the region that is being looked at. Data analysis parameters figure in as well. The higher % ID you pick OTUs at, the more species will be differentiated, for good or bad. example: two closely related species may be grouped together if you pick OTUs at 97%, where they may be classified as different OTUs at 99%.

Thanks a lot!