Application of Computer Aided Drug Design to Address Protein Flexibility in Ligand-Access Pathways and Inhibition of Adenylyl Cyclase Enzymes with Small Molecules

Neha Rana, Purdue University


Computational methods play a vital role in addressing difficult challenges in life sciences, including drug design and development. With the help of computational modeling, the molecular details of structural and functional aspects of biomolecules can be revealed. In this dissertation, two biological questions, prediction of ligand access pathways in enzymes with deeply buried active site and development of isoform-selective Adenylyl Cyclase inhibitors, have been dealt with computational techniques. A new computational modeling package has been developed, which can predict tunnels in enzymes with deeply buried active sites by including protein flexibility and protein-ligand interaction environment. The protocol utilizes coarse-grained representation of protein-ligand model and combines Elastic Network Modeling with Monte Carlo dynamics simulation to identify the course of ligand migration. A novel Elastic Network Model was designed, which specifically included ligand information as well as several new parameters that enhanced the performance of the model. Using the model, an ensemble of protein conformations was generated, which signified the long time scale dynamics of protein. The Elastic Network Modeling was then merged with an innovative protein-ligand Monte Carlo scheme. The Monte Carlo employed coarse-grained protein-ligand model, derived from predetermined protein and ligand conformations. The Monte Carlo implementation treated protein as kinematics chain of links and joints and handled an external library for sidechain rotamer conformations. In single MC trial step, a random conformation of protein was chosen and the ligand was transformed by a random translation and center of mass rotation, followed by random sidechain rotations in the proximity of the ligand. The perturbed move was examined for potential steric clashes using Bounding Volume Hierarchy (BVH). The move was scored using a coarse-grained protein-ligand statistical scoring function and accepted/rejected based on the Metropolis criteria. For prediction of tunnels, a geometric method based on Voronoi maps was used to identify the spaces for ligand traversal in the ensemble of protein conformations generated in ENM. The ligand was then sampled in small overlapping windows along the potential routes using MC technique. The MC technique was carried out for three consecutive windows before a new ensemble of protein conformation was rebuilt using ENM based on last location of ligand in the protein matrix. Any new potential routes were explored using MC technique. The process continued until the end of all routes. The sampling windows were assigned a harmonic potential, which were then pooled to derive Potential of Mean Force of each tunnel. For the second problem, well-established computational techniques were applied to understand mechanism of action of hits identified in high throughput screening against adenylyl cyclase isoforms. The 3D structures of adenylyl cyclase isoforms were derived using homology based modeling. The structures of protein and ligands were then used for binding site and binding mode identification using different molecular docking routines, such as Autodock4.2 and Glide. The binding poses and their binding free energy were confirmed with free energy calculations using molecular dynamics simulations. The mechanism of action and selectivity was deciphered using the protein-ligand dynamics and differences in the binding sites among isoforms.




Lill, Purdue University.

Subject Area

Molecular chemistry|Pharmaceutical sciences|Biophysics

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server