|advertisement: compare things at compare-stuff.com!|
The problem of representing protein structures accurately and intuitively in two dimensions has been tackled by a number of manual and automatic methods [Levitt & Chothia, 1976,Richardson, 1985,Hutchinson & Thornton, 1990,Flores et al., 1994,Barlow & Richards, 1995]. Any method, whether automatic or manual, requires the optimisation of the two-dimensional arrangement to preserve the `important' interactions at the expense of less important interactions. The assessment of importance often requires human intervention. The Kohonen mapping function, similar in broad principle to the Sammon mapping[Sammon, 1969] used by Barlow and Richardsbarlow:2d, can organise three-dimensional points in two dimensions preserving inter-residue distances as far as possible. However, unlike the Sammon mapping, data points are mapped to specific units in the map lattice and due to the preservation of local ordering, overlapping chain paths are discouraged and the representation is more idealised.