Homology-aware average nucleotide identity estimation at scale with TurboANI

Abstract

Here we introduce TurboANI, a fast homology-aware average nucleotide identity (ANI) estimation algorithm that implements the seed-chain-quasi-alignment scheme using SIMD seeding, compact indexing, lightweight chaining, and cache-friendly bitset sliding minwise hashing, achieving up to ~40,000-fold speedup over the alignment-based ANI algorithm with near-perfect agreement. TurboANI is also >40-fold faster than the state-of-the-art FastANI and ten-fold faster than skani while maintaining better accuracy. In the sensitive mode, it has a ten-fold higher true-ANI recovery rate than skani.