Chapter 6

Prediction of Loop Conformations in Antibodies

Abstract:

The Probability Grid Monte Carlo technique is an effective technique for many problems in protein modeling. Here, the technique has been adapted for modeling loop structures in proteins. The method has three phases: (1) generation of numerous loop backbone conformations using probability grids, (2) optimization of sidechain conformations of the best backbones from Phase 1 using probability grids, and (3) optimization of the best loop conformations from Phase 2 using conjugate gradients minimization. The method is applied to the six hypervariable loops of immunoglobulins, using the crystallized Fab fragments from HyHEL-5 and McPC603 as test cases. Conformations are predicted which are very similar to the crystal structure conformations - several backbone conformations have rms deviations of 1.0 Å or less from the crystal structure.