BAYESIAN CLUSTERING OF INDONESIAN RICE GERMPLASM

Model-based clustering where the inference on the parameters follow the Bayesian principle has been used to cluster 467 accessions of Indonesian rice germplasm which consist of released varieties, landraces, introduction lines, improved lines and wild species. A model-based Bayesian cluster analysis of genotype data can be used to evaluate the genetic backgrounds of rice populations of interest. Such analyses can be used to infer population structure, assign individuals to sub populations, and to study hybrid populations. Thus, the goal of this research was to examine the genotype data of numerous accession of rice germplasm using the model bayesian cluster analysis. The 1536 SNP-chip design was performed for genome scanning of the accession using the high throughput genotyping platform, the data of which were used for clustering. The result indicated that the germplasm can be clustered into five cluster based on similarities on genetic profile, i.e. similarities in gene frequencies across genome among individuals. Each cluster can be identified by reference lines, i.e. the lines or varieties that their genetic profile uniquely belong to one cluster and do not have or very rare introgression from lines or varieties of other clusters. Many introgressions have been identified among lines in all clusters which indicated that most of Indonesia rice germplasm, including local and introduced varieties were the results of crosses that occurred either in naturally fixation or breeding program activities that crossed one line/varieties to the others. There is also cluster in which no reference line and almost all lines/varieties in that cluster are known to have same common specific phenotype, e.g. aromatic.