Comparison of 16S- & WGS-derived community composition

Question

[lotus28]

Some studies point out that 16S and WGS-derived bacterial community compositions are significantly different.
But I have not seen any works actually comparing the accuracy of these 2 methods. What if we apply both of them to a gnotobiotic community? Which one will be more accurate?

While 16S produce more OTUs (possibly due to PCR chimeras), WGS might omit rare bacteria.
Both these methods produce protocol-dependent result, so even in a gnotobiotic experiment the true accuracy might be compromised by implementing a bad protocol for a possibly better method.

But what type of data should I use given the choice (e.g as in HMP that provides both 16S and WGS data)?

Answer 1

[5heikki]

Community compositions I have derived from 16S DNAseq, 16S RNAseq and WGS (from the same sample) have all differed significantly. I imagine contributing factors to this include at least varying 16S rDNA copy counts, 16S rDNA transcriptional activity, and genome sizes. However, what does it really matter? How much can you extrapolate from community composition with great confidence? If two genomes share a relatively similar 16S variable region they do the same thing? IMO 16S stuff is nice if you want to tell with relatively good confidence if something is there or not, but that's pretty much it.

Answer 2

[edward.messick]

I've run across a few studies that touch on this topic, I'll have to compile them and come back to this. But I've studied this myself for a while and WGS seems to be better for everything now, if implemented correctly. It hits species level specificity, whereas 16S is only sensitive at the genus level. WGS would actually be less likely to omit rare bacteria, in my opinion, due to the better databases available for genomes than 16S only. I'd love to continue this discussion, maybe see some dissenting opinions from mine. Again, I'll come back and post studies I've read because from what I remember they seem to contradict each other, and there still is some confusion on this topic overall.

[lotus28]

Thank you for the reply, I'll be waiting for you.

BTW, right now I'm trying to find compatible human gut metagenome data sets to study Crohn's disease (CD). So far, the HMP data looks big and good. It has 16S-metagenomes made with v1-v3, v3-v5 and v6-v9 primers and WGS-metagenomes. But CD-associated data have no WGS-metagenomes and their 16S-metagenomes are made with v1-v2 primers. Is it legit to compare communities derived from v13569 and v12 16S-sequencing?

If you could consult me on such details of metagenome sequencing, I'd like to contact you.

[edward.messick]

Sorry about the (very) delayed reply. After further study, 16S seems to be better for certain use-cases but I'd almost always suggest WGS at this point. There is also the major advantage of WGS being able to identify non-bacterial microbes, while 16S is limited to bacteria. Shoot me an email if you still need advice on this: [e-mail] (alternatively, if you are looking to buy services, I work for a genomics service provider doing bioinformatics work)