Résumé
Oryza sativa rice has been domesticated in Asia and has moved into Africa by various routes for over a millennium. Cultivars may have undergone a particular evolution during and after their migration, leading to novel adaptations. Making use of the resequencing data available in the 3000 Rice Genomes data set (Wang et al 2018), we applied a method recently developed (Santos et al., 2019) in order to relate the genetic diversity of this species between Africa and Asia. We used data from 1252 accessions, including all 249 accessions from Africa and 1003 landraces from Asia, characterised with over 10 million high quality SNPs. Our pipeline relies on Principal Component Analysis (PCA) for dimension reduction and on Kernel Density Estimation for discriminant analysis and classification of local genome segments. Local haplotypes are described in relation to Indica, Japonica and cAus references. This step is complemented by an analysis of local clusters for deeper exploration. We analysed over a hundred thousand overlapping windows of 150 SNPs. Summary analyses of local classification by genome and physical position allowed us to identify ambiguities between groups, which are possible signals of genetic exchange. Browsing through the genome, we detected potential introgressions between the different rice sub-populations and were able to isolate those specific to African accessions. We selected several on chromosome 6, corresponding to large (over 100 kb) segments of the Indica type or the cAus type observed in upland rice varieties globally classified as Japonica. We used these signals to start a systematic inventory of discrete introgressions in the rice genome and investigate their value to trace back the history of broad migrations and patterns of adaptation. Diversity in introgression-free genome portions indicates that African upland rice derives from Asian tropical Japonica rice with a significant decrease in diversity along the migration.