Comparison of Algorithms for Constructing Rice Core Collections

Core collection is a limited subset of accessions representing the spectrum of the whole collection with minimum repetitiveness. In this study, 96 rice accessions were sequenced by double digest Restriction Associated DNA sequencing (ddRAD) and resulted 1960 Single Nucleotide Polymorphism (SNP) markers. Methods for constructing core collections include random sampling, stratified sampling, Marita’s method, genetic distance sampling (GDOpt) and Core hunter (version 2.0). Average distance between each accession and nearest entry (A-NE) and the average distance between each entry and the nearest neighboring entry (E-NE) are used as criteria for evaluating the effectiveness of the methods tested. The results indicate that while GDOpt performed best in A-NE, Core hunter performed best in E-NE. However, core collections constructed by Core hunter favored accessions that are outbreeds between rice subpopulations and thus E-NE may not be an appropriate criterion when subpopulations were evident in the accessions. As the A-NE being the sole criterion, GDOpt is the method of choice for construction of core collection even when subpopulations exist as in the case of rice. The results of 413 diverse accessions based on the publicly available data of 44K SNP chip study also agree with the conclusion above.