*3.2. Population Structure and Genetic Relationships*

The model-based simulation of population structure analysis of the maize panel (208 accessions) showed that the delta K values from the mean log-likelihood probabilities plateaued at K = 3 (389.43), followed by K = 4 (276.33), and K = 2 (273.07) (Figure 1a). At K = 3, the 208 maize panel was divided into three sub-populations (Figure 1b). Using an 80% membership probability threshold, 122 accessions (58.65%) were successfully assigned to the three subpopulations. In comparison, 86 accessions with a probability of associations less than 80% were designated as an admixed population (Supplementary Table S2). Subpopulation 1 was the most uniform (membership coe fficient averaged, 90%), and it contained 53 landraces (49 from Burkina Faso, 3 from Togo, and 1 from Ghana). Subpopulations 2 and 3, which constituted 12.5% and 20.67% of the panel, respectively, were admixtures of Ghanaian and Togolese landraces, together with reference populations. Specifically, subpopulation 2 consisted of 26 accessions, 10 reference populations, and 11 and 5 Togolese and Ghanaian landraces, respectively. Subpopulation 3 comprised 24 and 15 landrace accessions from Ghana, and Togo, respectively, and 4 accessions from the reference population. The admixed group contained 60, 13, and 9 landraces from Togo, Ghana, and Burkina Faso, respectively, plus 4 reference populations (Supplementary Table S2). The additional smaller peaks observed at K = 4 (276.33) and K = 2 (273.07) implied the presence of subgroups within the three major groups (Figure 1). Therefore, an independent STRUCTURE run was performed for each gene pool. Sub-clustering of the Burkinabe and Ghanaian gene pools both yielded a sharp peak at K = 2 (Figure 2a,b). Sub-clustering the reference populations and Togolese landraces showed the highest peak at K = 3, and K = 9, respectively (Figure 2c,d). A substantial degree of admixture was observed for each gene pool (Supplementary Table S3).

**Figure 1.** Graphical representation of the population structure of the 208 maize panel. (**a**) Plot of mean likelihood of delta K against the number of K groups. The highest peak observed at K = 3 signifies the grouping of accessions into three groups, while the small peak at K = 2 and 4 signifies further grouping of accessions into two and four groups, respectively. (**b**) Subpopulations at K = 3. The colors represent three subpopulations of 208 accessions. The separation of accessions into subpopulation 1 (red), 2 (green), and 3 (blue) was based on membership coe fficient ≥80%.

**Figure 2.** Population structure of Burkinabe (**a**), Ghanaian (**b**), and Togolese (**c**) landrace gene pools, including a reference population (**d**) at K = 2, K = 9, and K = 4, respectively.

Using the Bayesian information criterion (BIC) implemented in DAPC, a maximum of K = 3 was obtained, which corresponded to three groups of maize accessions in the panel (Figure 3). Estimation of the cluster membership revealed that cluster three had the highest number of accessions (94) followed by cluster two with 77 accessions, and cluster one with the smallest number of accessions (37). Of the

94 accessions in cluster three, 58 (61.70%) and 29 (30.85%) were landraces from Togo and Ghana, respectively, including six reference populations and the landraces from Burkina Faso (Supplementary Table S4). All the accessions in cluster two were landraces from Burkina Faso (57), Togo (17), and Ghana (3). Of the 37 accessions in cluster 1, 14 (37.84%) were Togolese landraces, 12 (32.43%) were from the reference populations, while 11 (29.73%) were Ghanaian landraces.

**Figure 3.** Discriminant analysis of principal components (DAPC) using 5974 DArTseq markers. The axes represent the first two linear discriminants (LD). Each color represents a cluster, while each dot represents an individual. Numbers represent the different subpopulations identified by DAPC analysis.

Further investigation of the genomic structure of the maize panel using the PCoA indicated three subpopulations as per the STRUCTURE simulation and DAPC analyses (Figure 4). The total amount of genetic variation explained by the first two principal coordinates was 57%. The PCoA clearly separated subpopulation 3 (by PCo2), which showed a higher degree of admixture between Ghanaian and Togolese landraces, including six reference populations and a landrace from Burkina Faso. The other two subpopulations appeared to be distributed along PCo1. Although some degree of overlap among landrace gene pools was shown in subpopulation 1, located at the upper extreme of PCo1, ~75% were Burkinabe landraces. Subpopulation 2 distributed along the lower extreme of PCo1 was the most distant of the three, comprising the majority of the reference populations and four Ghanaian landraces.

**Figure 4.** Cluster analysis of 208 maize accessions using principal coordinate analysis (PCoA). Accessions are colored according to origin. The orange, ash, and violet circles represent subpopulation 1, 2, and 3, respectively.

As per the STRUCTURE, DAPC, and PCoA results, the NJ phylogenetic tree also showed three sub-populations with higher degrees of admixture among Ghanaian and Togolese landraces, and reference populations (Figure 5). The neighbor-joining tree performed for each gene pool divided the Burkinabe and Ghanaian gene pools into two main clusters (Figure 6b). The Togolese landraces and the reference populations were grouped into nine and three clusters, respectively (Figure 6b,c).

**Figure 5.** Phylogenetic tree estimated through the neighbor-joining method for 208 maize accessions from West Africa. The green, red, blue, and yellow clades and taxa represent Togolese, Burkinabe, Ghanaian, and reference populations, respectively. The orange, violet, and ash highlights represent subpopulations 1, 2, and 3, respectively.

**Figure 6.** Unrooted neighbor joining tree depicting genetic relationships among Burkinabe (**a**), Ghanaian (**b**), and Togolese (**c**) landrace gene pools, and a reference population (**d**). Colors of each tree represent genetic groups.

### *3.3. Analyses of Molecular Variance, Genetic Di*ff*erentiation, and Gene Flow among Gene Pools*

The AMOVA revealed that 14% of the total variation was found among gene pools, while the rest (86%) was within gene pools (Table 2). The overall *FST* value of the maize panel was 0.21, and the Nm value was 1.58. As shown in Table 3, the Burkinabe gene pool had the highest *FST* value (0.28), and the Ghanaian and Togolese gene pools had the lowest (0.18, each). The pairwise *FST* values ranged from 0.14 (Ghanaian vs. Togolese) to 0.31 (Burkinabe vs. reference populations). Similarly, Nm values between gene pools varied from 0.98 (Burkinabe vs. reference populations) to 2.83 (Ghanaian vs. reference populations). The Nm value between the Ghanaian and Togolese gene pools was 2.63.


**Table 2.** Analysis of molecular variance (AMOVA) using 5974 DArTseq markers of the genetic variation among and within four gene pools of 208 maize accessions.

df/degree of freedom, SS/Sum of square; MS/Mean sum of square, Est. Var./Estimated variance; Var./Variance.

**Table 3.** Measure of genetic population differentiation (*FST*) (lower diagonal), and estimation of gene flow (Nm) (upper diagonal) within and among the four gene pools of maize accessions.

