Whokaryote

Whokaryote classifies contigs from metagenomic assemblies as eukaryotic or prokaryotic using gene-structure features to improve domain-aware metagenomic analyses.

Key Features:

• Random Forest Classifier: Employs a random forest classifier using biologically grounded features such as intergenic distance, gene density, and gene length.
• High Accuracy: Achieves an estimated accuracy of 97% in contig classification and performs comparably to EukRep and Tiara, which rely on k-mer frequencies.
• Tiara Integration: Incorporates Tiara predictions as additional features to enhance classification, yielding an F1-score of 1.00 across precision, recall, and accuracy for both eukaryotic and prokaryotic classes.
• Computational Efficiency: Maintains fast processing speed appropriate for large metagenomic datasets.

Scientific Applications:

• Metagenome contig classification: Distinguishes eukaryotic and prokaryotic contigs in metagenomic assemblies to enable domain-specific downstream analyses.
• Discovery of eukaryotic biosynthetic gene clusters: Enables reanalysis of metagenomes to uncover eukaryotic genes and biosynthetic gene clusters, as demonstrated by detection of clusters in a disease-suppressive plant endosphere microbial community that were previously missed.

Methodology:

A random forest classifier is trained on genomic features (intergenic distance, gene density, gene length) and can incorporate Tiara predictions as additional features.