On 02/08/2012 12:47 PM, Mike Lawrence wrote:
Hi Francois
Here's a one-liner. The major concept behind the Sc coefficient is that it measures the
extent to which, "on average", the normal vectors between closest-neighbour
opposing points within the molecular interface are antiparallel.
Sc=1 implies that the surfaces fit exactly, all such vectors are perfectly
antiparallel. Heuristically, Sc values of 0.86 are about as good as
protein-protein interfaces get (see Nature. 2005 435, pp773-8). Values below
0.65 indicate relatively poor shape complementarity.
Use of a normal vector -based metric is considered superior to a distance-based
metric, though Sc does have a distance-based weight applied to the normal dot
products. Critical to this calculation is that the boundary of the buried
molecular interface has to be discarded from the measure, as this region is
intrinsically geometrically divergent. Sc is thus computed across only that
part of the buried surface that might be expected to be shape complementarity,
which makes it somewhat ill-suited to smaller interfaces.
All these details are in the JMB paper, which, unfortunately, there is no
substitute for reading :-)
Thanks a lot for this very nice answer.
I will definitely read your JMB article.
Best regards,
Francois.
