Comparison of differentiation estimates based on morphometric and molecular data, exemplified by various leaf shape descriptors and RAPDs in the genus Chaenomeles (Rosaceae)

Rumpunen, K.; Bartish, I.V.

Taxon 51(1): 69-82


ISSN/ISBN: 0040-0262
DOI: 10.2307/1554964
Accession: 003687197

Download citation:  

Article/Abstract emailed within 0-6 h
Payments are secure & encrypted
Powered by Stripe
Powered by PayPal

The efficiency of classical metric descriptors (MDs), normalized elliptic Fourier coefficients (EFCs), and the power series of normalized elliptic Fourier coefficients (PEFCs) in revealing variation in leaf shape was evaluated and compared for species in the genus Chaenomeles (C. japonica, C. speciosa, C. cathayensis and C. thibotica) using canonical variates analyses (CVAs) and reclassification tests. For EFCs, 30 harmonics (and for PEFCs 40 harmonics) were needed to achieve 100% correct reassignment of plants. By contrast, MDs were considerably less efficient and only 87, 77 and 66% of the plants were correctly reassigned to species, populations and maternal families, respectively. Furthermore, when compared with data based on molecular random amplified polymorphic DNA markers (RAPDs), PEFCs produced the most concordant data sets as revealed by cluster analyses. To obtain comparable estimates of variability and differentiation, the AMOVA approach was used both for molecular data and, as a novelty, for quantitative morphometric data. The variation was partitioned by hierarchical extraction of variance components from matrices of standardized squared Euclidean distances, and differentiation estimates were obtained as Phi -statistics. All descriptor sets partitioned the variance in a similar way. Surprisingly, high correlations were found between RAPDs and elliptic Fourier transforms for the among-family estimates of variance components (RAPDs vs. PEFCs: 0.89, P=0.003; RAPDs vs. EFCs: 0.86, P=0.006). In contrast, only a moderate correlation was found between RAPDs and MDs (0.71, P=0.049). This may indicate that shape per se (the size component excluded) is a less biased estimator of genetic variation than metric leaf descriptors in the genus Chaenomeles. Estimates of population differentiation based on EFCs and PEFCs were always lower than differentiation estimates based on RAPDs, whereas the differentiation estimates based on MDs were in general higher than estimates based on RAPDs, except for one species including hybrid populations.