Amplicon and metagenomics overview

Amplicon and Metagenomics Overview

This page presents a broad-level overview of amplicon sequencing and metagenomics as applied to microbial ecology. Both of these methods are most often applied for exploration and hypothesis generation and should be thought of as steps in the process of science rather than end-points – like all tools of science 🙂

Amplicon sequencing

Amplicon sequencing of marker-genes (e.g. 16S, 18S, ITS) involves using specific primers that target a specific gene or gene fragment. It is one of the first tools in the microbial ecologist’s toolkit. It is most often used as a broad-level survey of community composition used to generate hypotheses based on differences between recovered gene-copy numbers between samples.


Shotgun metagenomic sequencing aims to amplify all the accessible DNA of a mixed community. It uses random primers and therefore suffers much less from pcr bias (discussed below). Metagenomics enables profiling of taxonomy and functional potential. Recently, the recovery of representative genomes from metagenomes has become a very powerful approach in microbial ecology, drastically expanding the known Tree of Life by granting us genomic access to as-yet unculturable microbial populations (e.g. Hug et al. 2016; Parks et al. 2017).

Here we’ll discuss some of the things each is useful and not useful for, and then look at some general workflows for each.

Amplicon sequencing utility

As noted above, amplicon data can still be very useful. Most often when people claim it isn’t, they are assessing that based on things it’s not supposed to do anyway, e.g.:

“Why are you doing 16S sequencing? That doesn’t tell you anything about function.”

“Why are you measuring nitrogen-fixation rates? That doesn’t tell you anything about the proteins that are doing it.”

We shouldn’t assess the utility of a tool based on something it’s not supposed to do anyway 🙂

Metagenomics utility

With all that said, do you think we should expect relative abundance information from amplicon sequencing to match up with relative abundance from metagenomic sequencing?
No, and that's not a problem if we understand that neither are meant to tell us a true abundance anyway. They are providing different information is all. And the relative abundance metrics they do provide can still be informative when comparing multiple samples generated the same way 🙂

General workflows

Amplicon overview

PDF download

A Note on OTUs vs ASVs

All sequencing technologies make mistakes, and (to a lesser extent) polymerases make mistakes as well. Our initial methods of clustering similar sequences together based on percent similarity thresholds (generating Operational Taxonomic Units; OTUs) emerged as one way to mitigate these errors and to summarize data – along with a recognized sacrifice in resolution. What wasn’t so recognized or understood at first, is that when processing with traditional OTU-clustering methods, these mistakes (along with a hefty contribution from chimeric sequences) artificially increase the number of unique sequences we see in a sample (what we often call “richness”, though keep in mind we are counting sequences, not organisms) – often by a lot (e.g. Edgar and Flyvbjerg 2015, Callahan et al. 2016, Edgar 2017, Prodan et al. 2020).

Over the past decade, and particularly with greater frequency the past ~5 years, single-nucleotide resolution methods that directly try to better “denoise” the data (deal with errors) have been developed, e.g. Oligotyping (Eren et al. 2013), Minimum Entropy Decomposition (Eren et al. 2014), UNOISE (Edgar and Flyvbjerg 2015), DADA2 (Callahan et al. 2016), and deblur (Amir et al. 2017). These single-nucleotide resolution methods generate what we refer to as Amplicon Sequence Variants (ASVs). The field as a whole is moving towards using solely ASVs, and – in addition to being specifically designed to better deal with errors and successfully drastically reducing the number of spurious recovered sequences – there are pretty good practical reasons for this also. This Callahan et al. 2017 paper nicely lays out the practical case for this, summarized in the following points:

If you happen to work with amplicon data, and are unsure of what’s going on with this whole hubbub between OTUs and ASVs, I highly recommend digging into the Callahan et al. 2017 paper sometime as a good place to start 🙂

Metagenomics overview

PDF download

Some example tutorials



As you might guess, this is not as straightforward as the amplicon data tutorials as there are typically many more possible branching paths and things to chase with metagenomics data. But here are some places to start. If you know of others that are helpful please pass along so we can add them here 🙂