Understanding ANI: How Average Nucleotide Identity Defines Species

What ANI measures

Average nucleotide identity (ANI) compares two genomes base by base. One genome is cut into short fragments, typically about 1 kb, and each fragment is matched to the other genome. ANI is the mean identity of the fragments that match. Fragments with no match are left out, so every ANI value should be read together with its alignment fraction: the share of the genome that took part in the comparison.

The original method aligned fragments with BLAST (ANIb). FastANI estimates the same value with alignment-free approximate mapping and is up to three orders of magnitude faster, which makes it practical to compare one isolate against thousands of reference genomes ^[1].

Where the 95% threshold comes from

For decades, bacterial species were defined by DNA–DNA hybridisation: two strains with 70% or more DDH belonged to the same species. DDH is laborious and hard to reproduce, so microbiologists looked for a sequence-based replacement. Konstantinidis and Tiedje showed that whole-genome comparisons could advance the prokaryotic species definition ^[3]. Goris and colleagues then measured both values for the same strains and found that the 70% DDH cut-off corresponded to 95% ANI ^[2].

Richter and Rosselló-Móra confirmed that ANI mirrors DDH closely and placed the species boundary at about 95–96% ^[4]. The minimal standards for using genome data in prokaryotic taxonomy now recommend ANI, with this range, as the main genomic criterion for species delineation ^[5].

The threshold also reflects real biology rather than convention alone. When FastANI was run on all prokaryotic genomes in NCBI, 99.8% of the 8 billion genome pairs fell either above 95% ANI or below 83% ^[1]. Values between 83% and 95% are rare, which suggests that most bacterial species are separated by a genuine gap in genetic similarity.

How to read an ANI result

FastANI was designed for genomes in the 80–100% identity range ^[1]. Below that, too few fragments match and the estimate becomes unstable, so a low ANI should be read as "distant" rather than as a precise number.

Common pitfalls

• Contaminated or fragmented assemblies. A draft genome that contains reads from another organism can shift ANI and shrink the alignment fraction. Check assembly quality before interpreting taxonomy.
• Comparing against the wrong reference. ANI tells you how similar two genomes are, not what the species is called. The answer is only as good as the name attached to the reference genome, so compare against type strains where they exist.
• Ignoring the alignment fraction. A high ANI calculated over a small part of the genome can be misleading, for example when two genomes share a large plasmid but little else.
• Treating 95.0% as a hard line. Results between 95% and 96% need supporting evidence such as digital DDH or a core-genome phylogeny before naming a species.
• Assuming ANI is perfectly symmetric. Values can differ slightly depending on which genome is the query. Reporting both directions removes any doubt.

How ANI fits into a WGS report

In a BacSeq Analytics report, ANI is calculated with FastANI against the closest reference genomes after quality control and assembly. The report lists ANI and alignment fraction for each comparison, shows a matrix or heatmap when several isolates are analysed, and flags borderline results that need further checks. Species identification is included in every package, from Basic upwards.

For a probiotic or industrial strain, a correct species name matters beyond taxonomy. Safety assessments, regulatory submissions and comparisons with published strains all start from it.