Figure 3. The dbSNP build cycle starts with close of data for new submissions. We map all data, including existing refSNP clusters and new submissions, to reference genome sequence if available for the organism. Otherwise, we map them to non-redundant DNA sequences from GenBank. We then use map data on co-occurrence of hit locations to either merge submissions into existing clusters or to create new clusters. We then annotate the new non-redundant refSNP (rs) set on reference sequences and dump the contents of dbSNP into a variety of comprehensive formats on the dbSNP FTP site for release with the online build of the database.