Ultraconserved elements coupled with machine learning approaches resolve the systematics in model nematode species

Abstract

Nematodes are among the most diverse animals, yet only around 28,000 of an estimated one million species have been morphologically described. Their small size, morphological simplicity, and cryptic diversity complicate phylogenetic analyses. Traditional morphological and single-locus molecular approaches often lack resolution for both recent and ancient divergences. To address these limitations, we developed the first ultraconserved elements (UCEs) probe sets for two nematode families: Panagrolaimidae, a group of non-model organisms with limited genomic resources when compared to model taxa, and Rhabditidae, which includes the model species Caenorhabditis elegans. Our probe sets targeted 1,612 loci for Panagrolaimidae and 100,397 for Rhabditidae. In vitro testing recovered up to 1,457 loci in Panagrolaimidae, supporting robust phylogenetic reconstruction. Results were largely consistent with previous analyses, except for one strain reclassified as Neocephalobus halophilus BSS8. Using machine learning, we determined the minimum number of loci needed for accurate genus-level classification. For Rhabditidae, XGBoost achieved high accuracy with just 46 loci. For Panagrolaimidae, 39 loci were most informative. Our UCE-based approach offers a scalable and cost-effective framework for phylogenomics, enhancing taxonomic resolution and evolutionary inference in nematodes. It is well suited for biodiversity assessments and shallow, field-based sequencing, expanding research possibilities across this ecologically important phylum.