There are 14 independent DOMAK parameters (Table 6) for which suitable values had to be determined. Constraints on segment sizes (MSS) were derived by taking the smallest values of these constraints that appear in the set of domains that was derived by eye. MDS was chosen by looking at the sizes of domains in the same set. The smallest domain size in this set is actually 30, but this is exceptional so a size of 40 residues was chosen. MNCC values have not been optimised. sso was chosen to provide a compromise between speed and accuracy. If the amount of secondary structure in the segment is small, looking at secondary structure contacts only will not be accurate enough. The value was chosen by looking at two examples in which a sheet was being split (Brookhaven codes, 1PHA and 1IPD). HCD was set simply to the value of the average contact density in coil regions. MAC was derived by looking at the compactness of the domains that had been split by eye and choosing a value that encompassed most of them. MSV, MSVsso and MSVcs were derived by looking at the behaviour of five examples as these values were altered (1BBK (A chain), 1AMA, 1RHD, 1ALD, 1PHH). As with any analysis that categorises proteins on the basis of cut-off values, there are compromises made in choosing the cut-off values. The values found produce good results over the entire set, however, it may be possible to optimise them further.