Mining PDB with SPratt2
Take non-redundant subset of PDB, for example 800 chains.
Input all to SPratt2 in one go, search for patterns matching at least k structures and having maximum RMSd below <t, typical values: k=5, t=1.
Runs on a normal UNIX workstation in hours (or couple of days)
Thousands of patterns are produced, need post-processing step