Rarefaction Overview

​

What is Rarefaction?

Rarefaction describes how the number of features that are detected in a sequenced sample, whether distinct taxa, genes, or other units, changes as sequencing depth (i.e the number of sample reads) increases. A microbial community can comprise of hundreds to thousands of taxa (in complex environments), present at different abundances. If we extend the description of microbial profiles to functional genes, virulence factors or antimicrobial-resistance genes, the full feature space might extend well beyond thousands of features. In metagenomics, detecting a microbial species or genes encoding for an enzyme or AMR, regardless of the sequencing technology, requires enough reads originating from the corresponding genomic fragment in the sequenced sample. The probability of recovering enough reads to make a confident call on the presence of a feature depends mainly on two factors:

• Feature abundance in the community: Read counts scale with the relative amount of the corresponding DNA in the sequenced library, which depends on the abundance of the organism, its genome size and target copy number, and on how the sample was processed before sequencing.
• Sequencing depth ( i.e the total number of reads ): deeper sequencing samples more DNA fragments from the library and increases the chance of recovering enough reads from low-abundance taxa and genes. The depth required depends on the study goals: broad community profiling can be feasible at lower depth, whereas detecting rare taxa, AMR genes, differentiating between closely related strains requires deeper sequencing.

Rarefaction is a resampling analysis that repeatedly subsamples each sample’s feature-count table to a range of read depths and recalculates alpha diversity at each depth. It helps to asses:

• Whether, and to what extent, diversity estimates depend on sequencing depth
• How many samples would be retained at a chosen minimum read depth
• Whether differences between groups in downstream analyses may be driven by unequal sequencing depth rather than biological variation

​

Why would I do rarefaction?

​

Separating biology from sequencing depth

Differences in sequencing depth can arise during library preparation, sequencing, and downstream quality-control steps. Library preparation methods and sequencing platforms can introduce systematic technical variation, while high host-DNA content (such as low-biomass microbiomes, e.g: vaginal microbiome) can reduce the number of microbial reads available for microbial profiling after host read filtering.. Rarefaction helps separate biological and technical effects by showing how richness grows as more reads are included. Rarefaction curves let you see whether one group is consistently richer across a range of depths or only appears richer at higher sequencing effort, and, for a given sample type and taxonomic rank, they indicate depths beyond which few new taxa are detected. In complex ecosystems such as soil, some low-abundance taxa stay undetected at any realistic sequencing depth, because sampling never exhausts true sample richness. In practice, rarefaction curves are often used to choose a single rarefaction depth at which to normalize samples before calculating alpha diversity, beta diversity or performing other downstream analyses.

​

Was my sequencing depth enough?

A rarefaction curve that is still rising at your maximum sequencing depth means the sampling has not reached saturation: additional reads would likely reveal more taxa, or genes that remain undetected. This is a sign that your current depth may be insufficient to fully characterize the community, particularly for rare or low-abundance members. The depth at which a rarefaction curve begins to plateau is a practical, data-driven target for how much sequencing effort is needed to adequately characterize a given sample type. This is especially useful when planning for larger studies: rarefaction curves from a pilot dataset can help you estimate the sequencing depth required per sample before scaling up, balancing the cost of additional sequencing against the marginal gain in detected diversity.

​

How does it work?

The input to rarefaction is the feature table that describes the taxonomic or functional composition of your samples: the taxa or genes detected in each sample, given as read counts.