Nature Methods | Methagora

Is rare always real?

Deep sequencing of metagenomic datasets by next-generation technology, has revealed a richness and diversity of microbial species that far exceeded previous expectations.

In 2006 Mitchell Sogin presented the concept of a rare biosphere, low-abundance microbial populations that have persisted over large evolutionary time scales. Sogin described it as a ‘potentially inexhaustible reservoir of genomic innovation’ which lets microbial communities recover from environmental assault and allows them to adapt to changed circumstances.

How do scientists determine microbial diversity? Sequence reads covering the hypervariable regions of 16S rRNA genes are often used for classification. These ‘pyrotags’ (named thus since they are obtained on Roche’s 454 sequencer) can be classified by matching them against a rRNA sequence database.

The challenge is that for most microbial species the 16S rRNA gene is not known. An alternative is shotgun sequencing and grouping of reads into operational taxonomic units; i.e. a cluster of sequences with a defined difference to a neighboring cluster — commonly a sequence difference of 3% is required at the species level, 5% at the genus level.

But with predictions of an ever larger rare biosphere voices of caution are starting to be heard saying that its size is overestimated. Is everything that is classified as a new and rare species indeed real, or simply a sequencing error?

Most recently we featured an error correction program in our own pages that addressed this issue; and a note of caution by Phil Hugenholtz and colleagues about ‘Wrinkles in the rare biosphere’ just appeared in “Early View” in Environmental Microbiology.

How well founded are the concerns of an artificially inflated rare biosphere? Are current estimates of the rare biosphere really 10-fold too high? If so, what are the consequences — does it mean that the rare biosphere is far less important than assumed? Is it not such an inexhaustible reservoir after all but just background noise?

We understand that these are very controversial issues and we would love to hear from you.

Comments

  1. Report this comment

    Cliff Beall said:

    You have to feel that a lot of folks have egg on their face right now for rushing out 16S pyrosequencing publications without validating the system.

    I have noticed that a simple new noise reduction technique is to throw out any sequences that are seen less than 3 or 5 times. I have not seen that validated on known sequences, though.