*2.3. Data Analysis*

A total of 47,441 putative DArTseq markers were generated from the 208 maize panel. Prior to further analysis, the raw data set was filtered to remove markers with call rate <0.8, minor allele frequency (MAF) <0.05, and unmapped SNP markers. Thereafter, markers with no missing rate were retained using the TASSEL software version 5.2.12 [24]. The retained markers were subjected to various genetic diversity analyses including basic diversity statistics such as polymorphic information content (PIC), MAF, observed heterozygosity (Ho), and expected heterozygosity (He) using PowerMarker v. 3.2.5 [25]. The population structure of the maize panel was inferred using the Admixture model-based clustering algorithm implemented STRUCTURE 2.3.4 [26]. The *adhoc* number of clusters (k) was varied from 1 to 12, with 10,000 burn-in steps, followed by 10,000 Markov chain Monte Carlo simulations, as previously described [20,27]. For each k, ten independent iterations were implemented. The most likely number of k was determined by the ad hoc Δ k statistics [28] embedded in Structure Harvester [29]. Accessions with membership proportions (Q-value) ≥80% were assigned to groups, while those with membership probabilities less than 80% were designated as admixtures [30]. The population structure of each gene pool (Burkinabe, Ghanaian, Togolese, and reference populations) were also estimated as described above. A discriminant analysis of principal components (DAPC) was carried out on the 208 maize panel using the first 40 principal components using the adegenet R package [31]. Membership probabilities of the individuals for the different gene pools were estimated using the "find cluster" function implemented in adegenet. Further, principal coordinate analysis (PCoA) was conducted to reveal the genetic relationships among the maize accessions using GenAlEx v. 6.503 software [32]. An unrooted neighbor-joining (NJ) tree was constructed by following the procedure of Nei [33] with 1000 bootstrap replicates in PowerMarker v3.25 [25]. The resulting NJ tree was visualized in Molecular Evolutionary Genetics Analysis (MEGA) software version X [34] and edited using Figtree software v.1.4.4 [35]. Genetic relationships within each maize gene pool were elucidated through construction of an unrooted NJ tree, as described above. Analysis of molecular variance (AMOVA) was estimated in GenAlEx v. 6.503 [32] to partition components of genetic variance among and within the populations (k). Calculation of pairwise genetic differentiation statistics (*FST*) and haploid number of migrants (Nm) between gene pools was performed using GenAlEx v6.503 [32] with 999 permutations. *FST* measures the amount of genetic variance that can be explained by population structure based on Wright's F-statistics [36], while *Nm* = [(1/*FST*) − 1]/4. An *Nm* value less than 1 indicates limited gene exchange among subpopulations [36].
