At the current growth rate, global population is predicted to reach approximately 10 billion by 2050. As such, the demand for and pressure on resources, mainly land and water continue to increase with potential adverse effects from climate change posing further challenges. Moreover, with an increasingly urbanized world, the ratio of food producers to consumers has significantly declined. Under these conditions, societal consumption habits are also rapidly evolving and together, they pose a serious threat to global food security. To meet the resulting rise in demand for food, one requirement out of many other necessities is to increase agricultural production by over 50% without increasing the current agricultural land space. Crop management and breeding are essential pillars of effort to tackle the present and future challenges of food production byensuring yield growth to keep up with demand. Conventional breeding based on selection of genetic traits with best-performing qualities have accounted for about one-third of historical yield gains and other quality improvements. With the advancement in new technologies, molecular plant breeding application offers great promise for additional yield and crop quality improvement with DNA markers and marker assisted selection (MAS) tools. In addition, they are faster and come at a relatively lower cost. One approach of MAS, genome-wide association studies (GWAS), explores genetic variation in the whole genome to find signals of association between markers and the traits of interest. GWAS has been an effective tool in plant breeding however; it is mostly suited to the manipulation of few major genes. Unfortunately, most of the economically important traits that are crucial for the development of plant varieties are controlled by multiple small-effect genes. Genomic selection, a form of MAS has been designed specifically with the aim of improving quantitative traits especially those controlled by the many small genes. It uses genome-wide markers to predict breeding values of individuals as a base for selection in breeding programmes, without the need for identification of markers associated with traits of interest. The focus of this PhD thesis was to apply GWAS and genomic selection methods for seed quality traits assessment in elite winter barley originating from two commercial breeding programmes in Denmark. In the first study, a GWAS was conducted by estimating linear mixed model while adjusting for subpopulation stratification. Heritability estimates ranged from moderate (0.44) to strong (0.97). A total of 26 significant SNPs were identified for all traits except for inorganic phosphorous. The highly correlated traits like test weight and seed fractions were identified to be associated with the same significant marker. In addition, common SNPs located on chromosomes 4H and 5H were found to be associated with all the seed fractions. Relative to the high heritability estimated for the traits, the contribution of the SNPs in explaining the phenotypic variance wasquiet modest, however higher compared to other findings from the GWAS literature. In the second study, the two common approaches used in genomic selection (one-step and two-step approach) were compared using four prediction models (BRR, BayesA, BayesB, andBayesC). Three types of adjusted means mostly used in the two-step were also evaluated for their performance. Overall, prediction accuracies were high for both approaches in all traits ranging between 0.8 and 0.9. These high accuracies are mainly due to the high relatedness of the lines used in this study. There was no difference in the performance of the four models within the one-step and two-step approach. Moreover, the two-step approach did not show any significant difference from the one-step in terms of their prediction accuracies and means difference test between their GEBVs. Within the two-step approach, there was also no significant difference between the three adjusted mean types. This suggests that in the traits studied, the choice of model and approach should mostly depend on the researchers available resources since all methods arrive at the same results. In the final study, the aim was to evaluate pedigree-marker information and gene by environment (GxE) interaction effects on prediction accuracy using three-year dataset from the two breeding companies. Six mixed linear models where the main effect of lines, environments, markers, pedigree, and their interactions with environments were modelled using random covariance structures were analysed. Four cross validation schemes were also evaluated to predict within-company and across-company observations. The results showed that genomic selection models with only pedigree information performed similar to models with only marker information for most of the traits. Furthermore, models including marker and pedigree information jointly increased prediction accuracy compared to either marker or pedigree only models. Including GxE in the models showed an increase in prediction accuracy for all traits. It was also observed that the within-company predictions performed better than across-company predictions as expected. Nevertheless, with the degree of prediction accuracies obtained in the across-company cross validation, it is still possible to predict unobserved lines in one company with information from both companies.