Dental topography and diet in marsupials and comparisons with primates

Dental topography is a powerful tool for investigating the relationship between diet and tooth shape in mammals. Here, we test whether dental topographic metrics (DTMs) can accurately predict diet in marsupials based on 81 lower molars from 43 extant species in 12 families (six of the seven extant orders), and compare them to DTMs of primates with similar diets. DTMs tested included Dirichlet normal energy (DNE) and variants (ariaDNE, convex DNE, and ariaDNE CV), orientation patch count rotated (OPCR), and Relief Index (RFI). We also investigate the use of the ratio of trigonid to talonid height (TriTaHI) as an additional DTM. Notably, RFI and ariaDNE were less sensitive to sample composition and scanning modality than OPCR and ariaDNE CV were. For marsupials, highest dietary classification accuracy of the leave-one-out quadratic discriminant analysis was achieved by combining ariaDNE, RFI, molar size, and TriTaHI of lower second and third molars, resulting in 69.2% accuracy. In contrast, the primate sample of lower second molars reached highest classification accuracy (80.4%) when using only ariaDNE, RFI, and molar size. When using a clade-based cross-validated approach (i.e., using the primate sample to predict marsupial diets and vice versa), only ariaDNE, RFI, and molar size were needed to obtain the maximum classification accuracy (69.7% when classifying primates using the marsupial training set; and 61.8% vice versa). The comparative dataset of this study should prove useful for studies aiming to reconstruct the diets of fossil mammals that lack phylogenetically close extant analogues.