Identifying causal noncoding variants underlying complex traits in cattle remains challenging because high-resolution functional maps of regulatory variation are lacking. Here we combine massively parallel reporter assays with graph genomics to measure autonomous transcriptional activity from >1.5 billion DNA fragments spanning both cattle subspecies. In primary bovine cells, we assay >15 million variants and identify >150,000 expression-modulating variants enriched at cattle eQTL and GWAS loci. This enables the refinement of broad association signals to small sets of candidate functional regulatory variants. Our haplotype-aware framework captures rare, multi-allelic and tightly linked variants poorly resolved by conventional eQTL studies, and quantifies the disproportionate impact of larger variants on transcription. Furthermore, we use these data to train a deep-learning model that successfully predicts bovine promoter activity directly from sequence. Profiling the same cattle DNA in matched primary human cells reveals widespread conservation of promoter and enhancer activity, allelic effects and regulatory grammar, supporting the transfer of annotations and models across species. However, species-dependent effects are enriched in evolutionarily young sequences and p53-family motifs, highlighting the limits to simple cross-species extrapolation. Together, these data provide a high-resolution atlas of cattle regulatory variation and a framework for prioritising causal noncoding variants for cattle trait improvement. ### Competing Interest Statement LP, VB and JvA work for a commercial company, Annogen, that was paid to undertake aspects of this work. All other authors declare no competing financial interests. Biotechnology and Biological Sciences Research Council, BB/W000288/1, BBS/E/RL/230001B Gates Foundation, INV-076519 Department for International Development, https://ror.org/05rf29967 Roslin foundation, RF2301

Cross-species comparison and prediction of gene expression profiles are important to understand regulatory changes during evolution and to transfer knowledge learned from model organisms to humans. Si...

Genomes contain mosaics of discordant evolutionary histories, challenging the accurate inference of the tree of life. While genome-wide data are routinely used for discordance-aware phylogenomic analyses, due to modeling and scalability limitations, the...

Bayesian fine-mapping to obtain credible sets has been widely applied post GWAS to pinpoint causal v

Linking GWAS variants to their causal gene remains an ongoing challenge. A widely used method for pe

Identifying cell type-specific enhancers in the brain is critical to building genetic tools for inve

Bos taurus (taurine) and Bos indicus (indicine) cattle diverged at least 150,000 years ago and, sinc

Nature Methods - tidyomics offers a software ecosystem for omic data manipulation and analysis that bridges Bioconductor with the tidyverse framework.