Genomic and phenotypic characterization of finger millet indicates a complex diversification history

Advances in sequencing technologies mean that insights into crop diversification aiding future breeding can now be explored in crops beyond major staples. For the first time, we use a genome assembly of finger millet, an allotetraploid orphan crop, to analyze DArTseq single nucleotide polymorphisms (SNPs) at the sub-genome level. A set of 8,778 SNPs and 13 agronomic traits characterizing a broad panel of 423 landrace accessions from Africa and Asia suggested the crop has undergone complex, context-specific diversification consistent with a long domestication history. Both Principal Component Analysis and Discriminant Analysis of Principal Components of SNPs indicated four groups of accessions that coincided with the principal geographic areas of finger millet cultivation. East Africa, the considered origin of the crop, appeared the least genetically diverse. A Principal Component Analysis of phenotypic data also indicated clear geographic differentiation, but different relationships among geographic areas than genomic data. Neighbour-joining trees of sub-genomes A and B showed different features which further supported the crop’s complex evolutionary history. Our genome-wide association study indicated only a small number of significant marker-trait associations. We applied then clustering to marker effects from a ridge regression model for each trait which revealed two clusters of different trait complexity, with days to flowering and threshing percentage among simple traits, and finger length and grain yield among more complex traits. Our study provides comprehensive new knowledge on the distribution of genomic and phenotypic variation in finger millet, supporting future breeding intra- and inter-regionally across its major cultivation range. Core ideas ### Competing Interest Statement The authors have declared no competing interest. * Abbreviations : AMOVA : analysis of molecular variance; CV : coefficient of variation; DAPC : Discriminant Analysis of Principal Components; DArTseq : Diversity Arrays Technology sequencing; GRM : genomic relationship matrix; GWAS : genome-wide association study; LD : linkage disequilibrium; MAF : minor allele frequency; MLM : mixed linear model; MTA : markertrait association; NJ : neighbour-joining; PCA : principal component analysis; RR-BLUP : ridge regression-best linear unbiased genomic prediction; SNP : single nucleotide polymorphism