Significance Most sequence variant analyses rely on a linear reference genome that is assumed to lack millions of bases that occur in the genomes of other individuals. To quantify the extent and functional relevance of such missing bases, we integrate six genome assemblies from cattle and related species into a pangenome. This allows us to uncover more than 70 million bases that are not included in the Bos taurus reference genome. Through complementary bioinformatics, genomics, and transcriptomics methods, we discover putative genes from nonreference sequences that are differentially expressed and thousands of polymorphic sites that were unused so far. Our work provides a computational framework, broadly applicable to many species, to make a so-far neglected source of genomic variation amenable to genetic investigations.