Schematic showing Nearest-cell's logic. (1) The input cell is first converted to P1 if required. (2) It is then compared with every known P1 cell in the PDB using MATFIT (McLachlan, 1972; Kabsch, 1976, 1978); the schematic in box 2a shows an example superposition with one permutation of the database P1 cell (O′ superposed on O, A′ on A, B′ on B and C′ on C). If the lowest r.m.s. difference of all six superpositions is less than the specified cutoff (see §2.2.1), the database cell qualifies as a positive match. (3) The family-clustering algorithm clusters PDB entries into families of sequence similarity. Results are then displayed to the user with each family represented by the PDB entry with the smallest r.m.s. difference from the input. Families can be expanded to show all hits, as shown in Fig. 2.