ANI threshold for dereplication

Authors: orcid logoBérénice Batut avatar Bérénice BatutAdd Contributions!

Selecting an appropriate ANI threshold for de-replication depends on the specific goals of your analysis. ANI measures the similarity between genomes, but the threshold for considering genomes “the same” varies by application. Two key parameters influence this decision:

  • Minimum secondary ANI: The lowest ANI value at which genomes are considered identical.
  • Minimum aligned fraction: The minimum genome overlap required to trust the ANI calculation.

These parameters are determined by the secondary clustering algorithm used in tools like dRep ( Galaxy version 3.6.2+galaxy1)

Common Use Cases and Thresholds

  1. Species-Level De-Replication

    For generating Species-Level Representative Genomes (SRGs), a 95% ANI threshold is widely accepted. This threshold was used in studies like Almeida et al. 2019 to create a comprehensive set of high-quality reference genomes for human gut microbes. While there is debate about whether bacterial species exist as discrete units or a continuum, 95% ANI remains a practical standard for species-level comparisons. For context, see Olm et al. 2017.

  2. Avoiding Read Mis-Mapping

    If the goal is to prevent mis-mapping of short metagenomic reads (typically (( \sim )) 150 bp), a stricter 98% ANI threshold is recommended. At this level, reads are less likely to map equally well to multiple genomes, ensuring clearer distinctions between closely related strains.

Default and Practical Considerations

  • Default ANI threshold in dRep: 95% (suitable for most species-level applications).
  • Upper limit for detection: Thresholds up to 99.9% are generally reliable, but higher values (e.g., 100%) are impractical due to algorithmic limitations (e.g., genomic repeats). Self-comparisons typically yield (( \sim )) 99.99% ANI due to these constraints.
  • For strain-level comparisons, consider using InStrain ( Galaxy version 1.5.3+galaxy0) (Olm et al. 2021), which provides detailed strain-resolution analyses.

Important Notes

  • De-replication collapses closely related but non-identical strains/genomes. This is inherent to the process.
  • To explore strain-level diversity after de-replication, map original reads back to the dereplicated genomes and visualize the strain cloud using tools like InStrain (Bendall et al. 2016).

For further discussion on ANI thresholds, see Are these microbes the “same”? (Blog Post)

Still have questions?
Gitter Chat Support
Galaxy Help Forum

References

  1. Olm, M. R., C. T. Brown, B. Brooks, and J. F. Banfield, 2017 dRep: a tool for fast and accurate genomic comparisons that enables improved genome recovery from metagenomes through de-replication. The ISME journal 11: 2864–2868. 10.1038/ismej.2017.126
  2. Almeida, A., A. L. Mitchell, M. Boland, S. C. Forster, G. B. Gloor et al., 2019 A new genomic blueprint of the human gut microbiota. Nature 568: 499–504. 10.1038/s41586-019-0965-1
  3. Olm, M. R., A. Crits-Christoph, K. Bouma-Gregson, B. A. Firek, M. J. Morowitz et al., 2021 inStrain profiles population microdiversity from metagenomic data and sensitively detects shared microbial strains. Nature Biotechnology 39: 727–736. 10.1038/s41587-020-00797-0