Rapid and Accurate Estimation of Genetic Relatedness Between Millions of Viral Genome Pairs Using MANIAC

Abstract

AbstractThe estimation of Average Nucleotide Identity (ANI) plays a pivotal role in microbial and viral research, facilitating species delineation, taxonomy, genome dereplication in metagenomics and even detection of horizontal gene transfer. Traditional tools, optimised for bacterial genomes, fall short in addressing the complexities of phage genomics such as high sequence variability, mosaicism or the absence of universally shared genes. To bridge this gap, we introduce MANIAC (MMseqs2-based ANI Accurate Calculator), aiming to accurately estimate ANI and alignment fraction (AF) between pairs of viral genomes, using the MMseqs2 software which combines alignment-free and alignment-based approaches. We evaluated MANIAC against the gold-standard ANIb using complete phage genomes and further validated its performance with simulated and real genomic data. MANIAC demonstrated a near-perfect correlation with ANIb (R2= 0.999), outperforming existing tools like fastANI and Mash, especially for genomes below 80% ANI. When applied to hundreds of millions of pairs of phage genomes, MANIAC revealed a bimodal ANI distribution amongst phage populations, pointing to the existence of an ‘ANI gap’ similar to that observed in bacterial populations, albeit with quantitative differences. We then used a machine learning approach to classify same-genus pairs by combining both ANI and AF metrics, showing its strong predictive power (PR-AUC=0.970), particularly in virulent phages (PR-AUC=0.990). These findings underscore MANIAC’s potential to significantly advance viral genomics by providing a more accurate framework for quantifying genetic relatedness between viral genomes. MANIAC can be accessed underhttps://github.com/bioinf-mcb/MANIAC.