Population Structure and Association Mapping in Watermelon

Thirty five watermelons from the USA, Ukraine and Russia were evaluated for three seasons for their growth and fruit traits. The genotypic data generated by 196 SSR alleles and 1083 AFLPs was utilized to resolve population structure using STRUCTURE program. This analysis was conducted assuming two subpopulations (K=2) to eight subpopulations (K=8) using the SSR data and assuming two subpopulations (K=2) to eight subpopulations (K=8) using the AFLP data. The clustering results (six subgroups) were used as covariates in the association test (MLM procedure with TASSEL software). Combined means and correlations for various traits across the three growing seasons will be presented. Based on FST distributions across the clusters (Estimated using the program STRUCTURE2.3.2), we assumed that the watermelon cultivars in study have six sub-clusters. This analysis provided evidence on breeding histories based on the shared ancestries. LD (Linkage disequilibrium) was estimated separately for AFLPs and SSR data and marker associations and the corresponding p values will be presented. Our study concluded that there is significant moderate to high LD across the watermelon genome.  We performed association mapping using General Linear Model (GLM) and Mixed Linear Model (MLM) with shared ancestry (Q-matrix for both GLM and MLM and Kinship for MLM alone) using both sets of data (AFLP and SSR) across the three years (2005-2007) using the program TASSEL 2.1.  The P-value determines whether a QTL is associated with the marker and R² evaluates the magnitude of the QTL effects. Marker identification through MLM is known to be very robust and 2.5 times more powerful than the GLM. In the current study, a majority of QTLs showed that they are significant through the multiple years of evaluation. Common markers were identified for fruit length and width as well as soluble solids. This is the first QTL identification study using association mapping approach in watermelon. Best Linear Unbiased Prediction (BLUP) is a standard method for estimating random effects of a mixed model. We used BLUP to predict breeding values of watermelon cultivars taking into consideration of QTL genotype X environmental interaction. We conclude that in watermelon improvement programs, application of mixed models with random genetic effects can be very useful to estimate heritable genetic variance of various traits.