Flowchart illustrating the processing of input structures (ab initio decoys in this work) into MR search models and their subsequent trialling. Programs employed by AMPLE at various steps are given on the left. Steps not in italic text are carried out internally by AMPLE. The numbers to the right of the flowchart indicate typical numbers at different steps: each target will generate three SPICKER clusters, resulting in around 60 truncated ensembles and eventually the trialling of around 540 search models. On the right are snapshots during the successful solution of target 3hz7
. The largest SPICKER cluster consists of highly structurally diverse decoys, although a helix and following strands (less visible) are shared between them. The truncated ensemble in the same orientation shows greater structural homogeneity and has side chains placed with SCWRL, but still contains decoys that are too diverse to use as an ensemble MR search model. The search model shown results from subclustering, in this case using an r.m.s.d. threshold of 2 Å, and side-chain treatment; here, the retention of only more reliably predicted side chains. The behaviour of the top Phaser placement in SHELXE (an increase of the CC to a value of 49%) was indicative of success; indeed, the structure output by SHELXE (coloured blue to red from the N-terminus to the C-terminus) is almost complete and very largely correct compared with the deposited crystal structure (grey).