Figure 5.

Kraken database structure. Each k-mer to be queried against the database has a specific substring that is its minimizer. To search for a k-mer in the database, the positions in the database that contain k-mers with the same minimizer are examined. These positions are quickly found by examining the minimizer offset array for the start positions of records with the k-mer’s minimizer (orange) and the next possible minimizer (blue). Within a range of records associated with a given minimizer, records are sorted by lexicographical ordering of their k-mers, allowing a query to be completed by using a binary search over this range.