The dramatic increase in sequence information in the form of expressed sequence tags (ESTs) and genomic sequence has created a 'gene function gap' with the identification of new genes far outpacing the rate at which their function can be identified. The ability to create mutations in embryonic stem (ES) cells on a large scale by tagged random mutagenesis provides a powerful approach for determining gene function in a mammalian system; this approach is well established in lower organisms. Here we describe a high-throughput mutagenesis method based on gene trapping that allows the automated identification of sequence tags from the mutated genes. This method traps and mutates genes regardless of their expression status in ES cells. To facilitate the study of gene function on a large scale, we are using these techniques to create a library of ES cells called Omnibank, from which sequence-tagged mutations in 2,000 genes are described.