2.1. Genotyping and genetic clustering
For the present work, a total of 361 accessions of KG collected from Benin, Burkina Faso, Ghana and Togo were used. The DNA of each sample was extracted from young leaves of each accession, using the protocol of the Integrated genotyping service and support (IGSS) at the Biosciences Eastern and Central Africa (BecA: http://hub.africabiosciences.org/activities/services) located in Nairobi, Kenya. The quality of the DNA was confirmed by electrophoresis in 0.8% agarose, and the quantification carried out using UVP BioDoc-It2 Imaging System. All of the samples were diluted to 50 ng/µl for the DArT genotyping platform. Genotyping was performed using the DArT-Seq™ platform at Diversity Arrays Technology (Kilian et al. 2012). The quality analysis of the genotypic data was performed using Illumina HiSeq 2500 (Illumina 2015). The SilicoDArt calling algorithms (DArTsoft14) was used to score DArTseq markers into a binary format (presence = 1 and absence = 0) for each sample genomic representation. A total of 2844 SNP markers were obtained and processed in TASSEL v5 (Buckler et al.2014), for quality check. SNPs were filtered with TASSEL v5 for further analysis using the parameters Minor Allele Frequency (MAF) >0.05 and proportion of missing data < 30%. A total of 142 SNPs were retained and Nipals model in kdcompute (https://kdcompute.igss-africa.org) was used for data imputation.
The program Structure 2.3.4 (Pritchardet al. 2010) was used to assign individuals to different genetic clusters based on admixture model. The population structure was evaluated based on Bayesian clustering approach using the following settings: correlated allelic frequencies, burn-in period of 10,000 and 10,000 Markov Chain Monte Carlo (MCMC) interactions; and grouping (K) ranging from 1 to 5 in 10 independent runs. The results generated were used as input to the POPHELPER, an R package (Francis 2016), to predict the best K value based on Evanno method (Evannoet al. 2005) and to generate the bar graph and delta K graph.
The pairwise Fixation index (Fst) and molecular variation (AMOVA) were performed using R package adegenet (Jombart & Ahmed 2011) in order to reveal genetic differentiation among the two distinct groups occupying the three agroclimatic zones of Burkina Faso, Benin, Ghana and Togo. General patterns of genetic diversity were also evaluated by calculating observed (Ho) versus expected (He) heterozygosity and gene diversity (Hs) within each population.