SAS
  • Smooth S(X) by convolution with a Gaussian: S (X) = S(Y) h (X - Y) dY

    where h(X)= (22)-N/2 e-|X|**2 /2**2 ,  is SD of Gaussian, N is no. d.o.f.

    Because vdW potentials are very steep S (X) can practically not be calculated. This problem is overcome by the following transform from score space to probability space:

    S (X) = -RT ln (e-S(Y)/RT h (X - Y) dY )

  • Perform CG in a series smoothened landscapes with 1 > 2 > ... > n
Specs:

  • 150 cycles of max 6,000 steps of CG
  • initial = 3.0
  •  reduction factor between cycles: 0.93