Implementation of a Constrained Verlet Algorithm in NAMD and Molecular Dynamics Study of the BT Toxin Cyt1A


Xiaochuan Li

Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Chemistry and Biochemistry

First Advisor

Peter Butko

Advisor Department

Chemistry and Biochemistry


Cyt1A is a mosquito-specific toxin from Bacillus thuringiensis var. israelensis. NAMD, a high performance C++ object oriented parallel software, was used to simulate interactions between the Bt Toxin, water and the lipid membrane. Because the mutant Cyt1A (residue 225, Lys to Ala) lost its in-vitro and in-vivo activities, the study also included the interactions of the mutant toxin with water and the lipid membrane. We carried out all molecular dynamics simulations at constant temperature with periodic boundary condition for over 1 nanosecond. The study consisted of several conformations of Cyt1A with the membrane, inside the membrane and on top of the membrane. The MD simulation results identified the most flexible regions of the protein and showed that the mutant Cyt1A is more rigid than the native one. This indicated the mutant Cyt1A loses its cytolytic activities. We used CHARMM in the energy minimization of Cyt1A and the mutant Cyt1A. The minimization revealed that the conserved blocks are important in stabilizing the protein conformation. Electrostatic interactions played an important role in the toxin's function. The Adaptive Poisson-Boltzmann equation Solver (APBS) effectively mapped the electrostatic potential at the protein surface and was useful in the investigation of the energy difference between two conformations (Cyt1A and the mutant Cyt1A). With a constrained Verlet scheme in mind, we implemented a nonlinear Conjugate Gradient (CG) method in the NAMD. Our particular method contained strong Wolfe conditions. Constrained dynamics speeded up the simulations by a factor of two or three with small computational cost. The deduced matrix of linearized constrained equations was unsymmetric and crucial to numerical behavior of nonlinear CG. This M-matrix provided a basis for future algorithm development. These simulation results are in accord with the few available pieces of experimental data and will serve as a basis for future simulation studies.