Prediction of loop geometries using a generalized born model of solvation effects.

We have carried out an extensive exploration of the possibility of predicting the structure of long loops in proteins, using an 8- and a 12-residue loop in ribonuclease A as models. The native X-ray structure is used as a template while allowing for template flexibility; this makes our work relevant to the problem of homology modeling in which the template is not precisely known. Energies are calculated with the AMBER* and AMBER94 molecular mechanics potentials and the generalized Born continuum solvation model; and conformational space is sampled by means of a combination of Monte Carlo and molecular dynamics methods. Our AMBER94 results demonstrate that we can successfully generate loops with low root-mean-square deviations from the native as well as excellent energetic rankings.