Chandrajit Bajaj

Protein Protein Interaction

 Introduction | Representation | Function | Dynamic structure | Hierarchical model | Protein docking | Simulations | Meshing | Visualization

1. Introduction

Computational Modeling of Bionanomolecular Machines

Protein interactions with itself and other protein is an extremely important topic in computational biology. It helps solve many important questions including behavior of molecular structures important to the functioning of the human body and in the development of drugs. At CCV, we are trying to solve the important sub problem of handling flexible protein structures. Molecules in nature are seen to be flexible and capable of folding in solvents, giving it distinctive characteristics.

Once we have a flexible protein model, we would like to simulate the behavior of the molecule in solvents, and in the presence of other molecules. Problem docking is a specific problem in this domain. Here, we are interested in trying to apply the physics and chemistry behind the interactions and simulate the process. 

Visualization of the entire process is an important tool for the following reasons

  • Scientific discovery is enhanced by providing users with constant visual and other feedback from the simulations. Identification of behavior and potential changes over space and time is useful to understand the simulation
  • Verification and recalibration of simulations can be better performed through visualization.

Some of the issues that comes up in flexible protein dynamics are

  • Flexible model representation
  • Animation of the simulation at interactive speeds
  • Representation and rendering of both surfaces of interest and volumes or regions of the fields where the interactions occur.
  • Hierarchical data structures for proteins described by the PDB
  • Sparse formats to reduce both time and speed complexity of protein docking. Previous methods which include fourier, spherical harmonic and wavelet representations are seen to be extremely space and time consuming and often without specific error bounds.