Parallel Computations for Macromolecular Simulation

1
Parallel Computations for Macromolecular
Simulation

• Florida International University
• Department of Chemistry and Biochemistry
• Cassian DCunha
• Dr. David Chatfield

2
Importance

• Computer simulations of biological macromolecules
  are widely used
• Pharmaceutical industry to accelerate drug
  discovery
• Academic research for a host of problems from
  protein folding to enzyme mechanism

3
Issue With Macromolecular Simulations

• The sizes of the molecular systems are very large
  and the time scales modeled are memory and
  compute time intensive
• Time scales of possible simulations is directly
  related to the computer resources available

4
Introduction

• Our research is focused on an enzyme,
  chloroperoxidase (CPO), of industrial interest as
  a potential biological chiral catalyst
• Isolated from the mold Caldariomyces fumago
• Heme protein

5
Chloroperoxidase (CPO)
6
Importance of CPO

• Halogenation
• Normal function
• Dehydrogenation
• Heme peroxidase
• H2O2 decomposition
• catalase
• Oxygen insertion
• P450s
• Epoxidation

7
Research Goal

• Our research is aimed at elucidating the reaction
  mechanism of CPO and at predicting useful
  site-specific mutations for fine-tuning the
  enzymes catalytic activity for particular
  substrates of interest.

8
Halogenation Reaction
9
Computations

• The calculations we need to perform are of two
  types, both of which require substantial computer
  power
• Quantum mechanical (QM) calculations for the
  mechanism work
• Molecular mechanical (MM) calculations for the
  mutation work

10
Resource Available for Research

• A computer cluster with thirty dual-processor AMD
  Athlon nodes
• Software used for our research are gaussian98,
  CHARMM and Turbomole

11
Cyberbridges

• Reconfigure the computer cluster (ROCKS)
• Installation and compilation of parallel versions
  of the software used
• Benchmark computations carried in parallel mode
  and compare with those run in serial mode
• Optimal use of the hardware available
• Collaborate with students in different research
  areas

12
Test Results

• Parallel implementation of CHARMM using MPICH
• Short dynamic simulation of CPO for comparison
  using 1, 2, 4, 8, 16 CPUs
• The communication between nodes seems to be the
  bottleneck as the number of processors used is
  more than two

13
CHARMM
14
Future Work

• Implement Turbomole and gaussian98 in parallel
• Implement QM/MM calculations using the ONIOMM
  (Own N-layered Integrated Orbital and Molecular
  Mechanics) methodology

15
ONIOMM
16
Acknowledgement

• The CyberBridges staff
• The CyberBridges project that has supported our
  research has been funded by the NSF award
  0537464