HaloClassifier: integrating coding-signature features and k-mer composition for plasmid-chromosome discrimination in haloarchaeal genomes
Background: Plasmids are key drivers of horizontal gene transfer (HGT), enabling the dissemination of accessory traits that shape microbial adaptation and ecological interactions. In Haloarchaea-dominant members of hypersaline environments-characterizing plasmidomes remains particularly challenging because most available genomes are incomplete, leaving many contigs unassigned to either…
HaloClassifier is a novel machine-learning tool designed to differentiate plasmid and chromosomal contigs within haloarchaeal genomes. Unlike existing bacterial-centric methods, HaloClassifier uniquely combines haloarchaeal-specific genomic signatures with coding-derived features, which have been underutilized but highly informative predictors. The model also incorporates a variable-selection strategy that significantly reduces model complexity and computational cost without sacrificing predictive accuracy.
When trained on simulated contigs derived from complete haloarchaeal genomes, HaloClassifier achieves impressive accuracy rates. On extralarge contigs (40-100 kb), the model attains 96.61% accuracy at a 0.6 classification threshold, classifying 3.55% of contigs as unclassified. For smaller contigs (1-5 kb), HaloClassifier maintains 79.51% accuracy overall, with 84.42% classification at the 0.6 threshold and 17.07% of contigs remaining unclassified.
The tool's robust performance extends to metagenomic datasets, filling a critical gap in the field of Haloarchaea research. By providing accurate plasmid identification in fragmented assemblies, HaloClassifier lays the foundation for large-scale plasmidome reconstruction in these hypersaline microorganisms. This framework will facilitate future investigations into gene mobility, ecological adaptation, and the evolutionary dynamics of plasmids and horizontal gene transfer within Haloarchaea.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
