soibean: High-resolution Taxonomic Identification of Ancient Environmental DNA Using Mitochondrial Pangenome Graphs

Abstract

Ancient environmental DNA (aeDNA) is becoming a powerful tool to gain insights about past ecosystems. However, several methodological challenges remain, particularly for classifying the DNA to species level and conducting phylogenetic placement. Current methods, primarily tailored for modern datasets, fail to capture several idiosyncrasies of aeDNA, including species mixtures from closely related species and ancestral divergence. We introduce soibean, a novel tool that utilises pangenomic graphs for identifying species from ancient environmental mitochondrial reads. It outperforms existing methods in accurately identifying species from multiple sources within a sample, enhancing phylogenetic analysis for aeDNA. soibean employs a damage-aware likelihood model for precise identification at low-coverage with high damage rate, demonstrating effectiveness through simulated data tests and empirical validation. Notably, our method uncovered new empirical results in published datasets, including using porpoise whales as food in a Mesolithic community in Sweden, demonstrating its potential to reveal previously unrecognised findings in aeDNA studies.