NeighbourGroups: a machine learning classification tool that assigns microbial multi-locus genotypes to clusters

Robust microbial classification systems are essential, but their definition is complicated by the large size and high diversity of microbial populations combined with a widespread horizontal genetic exchange. Multi-locus approaches that index gene variation without explicit phylogenetic classification mitigates these problems, but reproducibly defining high-level groups remains problematic. We describe a generalisable machine learning approach, 'NeighbourGroups', that reproducibly, robustly, and rapidly classifies multi-locus sequence types with defined precision.