Computational studies of protein-ligand interaction through QM/MM methods and virtual screening

Computational methods have been successfully used to speed the drug discovery process and to provide a more robust search of chemical space for new drug candidates. These goals are accomplished through calculation of free energy upon binding using various methods or the use of virtual screening. Both methods can be used to theoretically predict a ligand’s effectiveness before proceeding with the more costly experimental phase. This dissertation will discuss the development and use of these methods and their applicability to a drug discovery application.;The first two chapters of this dissertation will provide a background for computational techniques used in drug discovery. The techniques presented range from knowledge-based scoring functions from a statistical analysis of the PDB to using quantum mechanical parameters from a semi-empirical method. These chapters will also provide information on mixed Quantum Mechanical/Molecular Mechanical (QM/MM) methods as well as calculating important terms when computationally determining binding free energies such as solvation and entropy.;The next two chapters detail the development and implementation of a QM/MM solvation model and scoring function. DivCon was integrated with the AMBER molecular mechanics program to produce a QM/MM package that can perform linear scaling semi-empirical calculations using AMBER’s force fields. The solvation model is based on a finite-difference Poisson-Boltzmann solver available in the DivCon program. This method was validated through calculation of solvation energy on a set of pentapeptides as well as small proteins. The QM/MM solvation energy was compared to the full QM solvation energy for various sized QM regions and only small differences were found. The QM/MM scoring function made use of this solvation method to calculate the solvation free energy of 23 zinc containing proteins. This QM/MM scoring function was validated against the set of proteins by comparing the predicted binding free energies of these proteins to the experimental binding free energies.;Following this, the next chapter discusses the use of virtual screening methods to find a new inhibitor of a deubiquitinating cysteine protease protein, HAUSP/USP7. DOCK and AutoDock were utilized to screen a commercially available subset of compounds from the ZINC database. First, a potential active site was identified, and then the ligands were docked and ranked. These ranked compounds were then analyzed, purchased and tested experimentally to determine their potency for the target protein. Several ligands bound with low micromolar efficiency and the results were examined. Future directions for this study are also suggested at the end of this chapter to determine a more specific stronger binding ligand.