Selection of the most discriminating morphological qualitative variables for characterization of fig germplasm

Abstract

Most descriptor lists for the characterization of genetic resources in plants include a large number of traits whose evaluation is a lengthy and expensive process making the characterization of large germplasm collections difficult. Consequently, to facilitate the study and the conservation of germplasm, it is important to select carefully the most informative variables for each species. In this work, we applied sequential statistical procedures to select the most discriminant variables in fig (Ficus carica L.) from the initial 134 qualitative variables studied. A total of 34 variables was finally selected and broken down in 97 characters that were grouped by principal component analysis in 11 principal components that explain 93.34% of the total variability. The unweighted pair group method with arithmetic mean dendrogram derived from a similarity matrix generated using the Pearson's correlation coefficient among the 11 principal components selected classifies the cultivars in four main groups mainly based in the production type. These results show that redundant information can be obtained in morphological studies with a large number of variables resulting from correlation between variables. Consequently, a carefully selected and reduced number of highly discriminating variables can allow efficient fig germplasm characterization and discrimination resulting in significant savings of time and resources.