This thesis presents a novel method called meta-selection-scores (MSS) to construct a meta-assembly of genome-wide signatures of selection within breeds (n = 24) and across geographical archetypes (European, African, Zebu and composite). The meta-assembly and review of published studies highlight the historical selection events, the role of underlying genes for various traits, limitations of the available bovine genomic resources and implications of using different methodologies. This study also developed another new method called composite selection signals (CSS) – to improve the power of selection signature scans – in which multiple pieces of evidence for selection derived from the rank distribution of individual tests are combined in a single score. CSS has shown improved power by localizing the genomic regions under selection for major traits in multi-breed panels of cattle and sheep. CSS was also used to detect regions associated with the complex trait of bovine stature and implicated 12 (nine were novel) regions harbouring multiple candidate genes in contrasting cohorts within the European and African cattle. Without phenotypic records on individual animals, the CSS method provides a firsthand scan of the genome to detect putative regions associated with complex traits. The genome-wide CSS scans for individual breeds (n = 60) of cattle using previously available and new genotypes from 50K SNP chip assay identified 177 genomic regions under selection. Finally, an ultra-high density dataset (1.6 million SNPs) was investigated with CSS for Angus and Holstein. A high (8 of 9) reproducibility of the 50K SNPs based CSS regions and up to 14 fold additional genomic regions under selection were detected using 1.6 million SNPs. Overall, this study provides a detailed investigation about the core traits influenced by the historical selection events in worldwide cattle breeds and presents novel insights about the hotspots of positive selection in the bovine genome.