Fig 1.
Conventional banana breeding starts with crossing 3x inferior and parthenocarpic landrace varieties (A) with a wild, seeded 2x accession (B). 4x resulting from this cross (C) are selected and crossed with improved 2x hybrids (D). The resulting secondary 3x (E) are selected and evaluated as potential improved varieties. This process takes up to 15 years.
Table 1.
Pearson’s correlation coefficients of traits under low input field management (GS1).
Table 2.
Pearson’s correlation coefficients of traits under high input field management (GS2).
Fig 2.
Principal component analysis plots generated in R using package FactoMineR for the traits scored in a banana genomic selection training population.
(A) shows the distribution of traits under low input field management (GS1) and (B) shows the distribution of traits under high input field management (GS2) on the first two components.
Fig 3.
Principal component analysis plots generated in R using package FactoMineR for the cross combinations in a banana genomic selection training population.
(A) shows the distribution of cross combinations under low input field management (GS1) and (B) shows the distribution of cross combinations under high input field management (GS2) on the first two components.
Fig 4.
Principal component analysis plots generated in R using package FactoMineR for the traits scored in a banana genomic selection training population.
(A) shows the distribution of traits for individual genotypes and (B) shows the distribution of individual genotypes on the first two components based on mean of combined data from the two fields.
Fig 5.
Effect of cycle on trait variation in bananas, where (a) shows an increase in plant height at flowering at cycle 2 while (b) shows no increase in index of non-spotted leaves at cycle 2.
Table 3.
Coefficient of determination and Student’s t-test P-values showing the effect of cycle on cross combinations.
Table 4.
Effect of genotype (clone), field management, cycle and their interaction on trait variation.
Table 5.
Comparison of mean performance of parental cross combinations (S) and hybrids from those crosses (R) against the mean of all parents.
Fig 6.
Dendrogram showing the genetic diversity of the genomic selection training population based on 19 informative SSR markers.
The squared Euclidean distances were used to generate the hierarchical clusters based on ward.D criterion. Where cluster A = tetraploids (4x) by M. a. spp. malaccensis 250, * share only female parent, cluster B = matooke (EAHB), cluster C = tetraploids from EAHB (3x) by Calcutta 4 a wild diploid (2x), cluster D = wild and improved diploids, cluster E = Black Sigatoka resistant diploid hybrids, cluster F = hybrids of 5610S-1 as a male parent, * share grandparent Calcutta 4, GC = good for cooking and N = NARITA hybrid, cluster G = cv. Rose was the main male parent, * share genetic background, cluster H = Long Tavoy and Calcutta 4 are the grandparents, cluster I = mostly hybrids of SH3217 as male parent, N = NARITA, @ = released variety as NARITA 7/M9/Kiwangazi and cluster J = triploid hybrids with complex pedigree, most advanced hybrids such as NARITAs (N) are found in this cluster of which some are promising variety candidates and GC = good for cooking.