Résumé
The principal component score strategy (PCSS) is a multivariate method which allows the identification of a core subset from a germ-plasm collection. Previously described for quantitative data, the method is extended here to qualitative data provided by molecular markers. Quantitative and qualitative PCSS were then applied to real data on four tropical crops: coffee, rice, rubber tree and sorghum. The results show, in all cases, that the increase in the cumulated relative contribution (CRC) is very rapid but may depend on the species. Ten percent of the entire collection yielded between 22 and 58 % of the CRC. As expected, the variability of the quantitative characters in the subsets was little or not modified by a qualitative selection but largely increased by a quantitative one, whereas qualitative PCSS was more efficient in preserving rare alleles and increased the global diversity with limited quantitative changes. The range of crop plants tested made it possible to compare the respective impacts of the two methods and highlighted the advantage of combining both types of characters. (C) Inra/Elsevier, Paris