Comparing population patterns for genetic and morphological markers with uneven sample sizes: An example for the butterfly Maniola jurtina

Integrating genetic and/or phenotypic traits at population level is considered a fundamental approach in the study of evolutionary processes, systematics, biogeography and conservation. But combining the two types of data remain a complex task, mostly due to the high, and sometimes different, sample sizes required for reliable assessments of community traits. Data availability has been increasing in recent years, thanks to online resources, but it is uncommon that different types of markers are available for any given specimen. We provide new R functions aimed at directly correlating traits at population level, even if data sets only overlap partially. The new functions are based on a modified Procrustes algorithm that minimizes differences between bidimensional ordinations of two different markers, based on a subsample of specimens for which both characters are known. To test the new functions, we used a molecular and morphological data set comprising Mediterranean specimens of the butterfly Maniola jurtina. By using this method, we have been able to maximize similarities between genotypic and phenotypic configurations obtained after principal coordinate analysis for the model species and evaluated their degree of correlation at both individual and population level. The new recluster.procrustes function retained the information of the relative importance of different morphological variables in determining the observed ordinations and preserved it in the transformed configurations. This allowed calculating the best combination of morphological variables mirroring genetic relationships among specimens and populations. Finally, it was possible to analyse the modality and variance of the phenotypic characters correlated with the genetic structure among populations. The genetic and phenotypic markers displayed high overall correlation in the study area except in the contact zone, where discrepancies for particular populations were detected. Interestingly, such discrepancies were spatially structured, with southern populations displaying typical western morphotype and eastern haplotypes, while the opposite occurred in the northern populations. The methodology here described can be applied to any number and type of traits for which bidimensional configurations can be obtained, and opens new possibilities for data mining and for meta-analyses combining existing data sets in biogeography, systematics and ecology.