Background: Parasites cause predictable alternative phenotypes of host individuals. Investigating these parasitogenic phenotypes may be essential in cases where parasitism is common or taxa is described based on a parasitized individual. Ignoring them could lead to erroneous conclusions in biodiversity-focused research, taxonomy, evolution, and ecology. However, to date, integrating alternative phenotypes into a set of wild-type individuals in morphometric analysis poses extraordinary challenges to experts. This paper presents an approach for reconstructing the putative healthy morphology of parasitized ants using algorithmic processing. Our concept enables the integration of alternative parasitogenic phenotypes in morphometric analyses.
Methods: We tested the applicability of our strategy in a large pool of Cestoda-infected and healthy individuals of three Temnothorax ant species (T. nylanderi, T. sordidulus, and T. unifasciatus). We assessed the stability and convergence of morphological changes caused by parasitism across species. We used an artificial neural network-based multiclass classifier model to predict species based on morphological trait values and the presence of parasite infection.
Results: Infection causes predictable morphological changes in each species, although these changes proved to be species-specific. Therefore, integrating alternative parasitogenic phenotypes in morphometric analyses can be achieved at the species level, and a prior species hypothesis is required.
Conclusion: Despite the above limitation, the concept is appropriate. Beyond parasitogenic phenotypes, our approach can also integrate morphometric data of an array of alternative phenotypes (subcastes in social insects, alternative morphs in polyphenic species, and alternative sexes in sexually dimorphic species) whose integrability had not been resolved before.
Keywords: Temnothorax; cysticercoid; machine learning; morphometry; parasitogenic phenotypes.