Semester

Fall

Date of Graduation

2022

Document Type

Thesis

Degree Type

MS

College

Eberly College of Arts and Sciences

Department

Chemistry

Committee Chair

Blake Mertz

Committee Co-Chair

Hacer Karatas Bristow

Committee Member

Werner J. Geldenhuys

Committee Member

Mark L. McLaughlin

Abstract

Docking simulations are a vital virtual screening technique to identify potential drug candidates by predicting the binding score and position for protein-ligand interactions. As ligand libraries expand into the billions, the number of groups able to run these simulations decreases drastically. Here we propose a simple methodology allowing for more accessibility of docking large ligand libraries. This method uses simple molecular features and unsupervised learning to cluster like molecules. These clusters are then randomly sampled, creating a training set. A model is trained using the bag of bonds method and a simple neural network to predict the binding affinity of the rest of the library. Using a ten thousand class A, G-protein coupled receptor (GPCR) ligand library, the proposed method is tested against two conformations of the protein platelet-activating factor receptor (PAFR). We could quickly and accurately predict the binding affinity using only five percent of the library through this method.

Embargo Reason

Publication Pending

Share

COinS