This thesis examines various methods for generating multi-trait polygenic risk scores (PRS). The primary objective is to see which multi-trait method performs best and are there any simpler methods that can perform as well. The thesis evaluates each method by comparing the weighted-average multi-trait PRS with true phenotype values (target traits), using the correlation coefficient (ρ) for continuous traits and the area under the receiver operating characteristic curve (AUC) for binary traits. It also investigates how different simulation parameters influence performance. Two additional novel multi-trait PRS methods are introduced in this work: mt-lm and mt-CVb. mt-lm is essentially a multiple linear regression for a continuous focal trait and logistic regression for a binary focal trait, while mt-CVb combines cross-validation and bagging techniques in a hybrid approach to improve model performance. The existing multi-trait method wMT-SBLUP consistently achieves the best performance, outperforming all other methods in most scenarios. While the two novel methods are not the top performers, they demonstrate better results compared to other methods (excluding wMT-SBLUP) for both continuous and binary focal traits across various parameter settings. Moreover, mt-lm offers the additional advantage of being faster than wMT-SBLUP. / Thesis / Master of Science (MSc)