iDiv Annual Conference 2025

Testing machine learning approaches for species delimitation in taxonomically complex groups – A case study in Toona (ENDL.) M.ROEM., Meliaceae

Not scheduled

1m

Lecture Hall XXII (Audimax)

Poster Molecular Biodiversity and Evolution Molecular Biodiversity and Evolution

Speaker

Jana Ebersbach (Department of Molecular Evolution and Plant Systematics & Herbarium (LZ), Institute of Biology, Leipzig University, Leipzig, Germany; German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, Leipzig, Germany)

Description

The ongoing biodiversity crisis necessitates rapid and comprehensive assessments of global species inventories which is hampered by the slow pace of species discovery and formal taxonomic description. Recent advances in machine learning hold much promise for fast-tracking species delimitation and integrative taxonomic approaches yet their applicability in taxonomically complex groups remains underexplored. The small but socioeconomically important tropical timber genus Toona. (4-6 species, Meliaceae A. JUSS.) has thus far evaded taxonomic clarification and thus represents an ideal system to test these new techniques. For this project, we will compare established species delimitation approaches (e.g., Stacey, DAPC) to unsupervised machine learning (SuperSOM) and use the resulting genetic clusters as labels to train a Support Vector Machine classifier using image scans of the sampled Toona herbarium vouchers. Ultimately, this approach promises to unlock the diverse information available from herbarium resources and pave the way towards resolving other taxonomically complex groups.