Electronic Theses and Dissertations

Date

2021

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Chemistry

Committee Chair

Nathan DeYonker

Committee Member

Abby Parrill-Baker

Committee Member

Henry Kurtz

Committee Member

Ph.D.

Abstract

In order to accurately simulate the inner workings of an enzyme active site with quantum mechanics (QM), not only must the reactive species be included in the model, but also any important surrounding residues, solvent, ions, and coenzymes involved in crafting the microenvironment. The Residue Interaction Network ResidUe Selector (RINRUS) toolkit was designed to utilize interatomic contact network information for automated, rational residue selection and QM-cluster model generation. An X-ray crystal structure of a protein is translated into a two-dimensional network which may be then used to discern residues with significant interactions with the enzyme substrates. The rest of the protein is trimmed away following a defined protocol to create QM-cluster models suitable for simulation.Three QM-cluster enzyme case studies demonstrating the capability of network-based models are presented in this work. First, models of six bioengineered threonyl-tRNA synthetase enzymes are simulated to reveal the impact residue mutations have towards creation of a transition state analogue structure within a protein pocket. Second, models of the zinc-native enzyme human carbonic anhydrase II with various transition state ions in the active site are shown to provide insight into the reduced catalytic activity of the metallovariants, along with predicting the potential viability of the iron-substituted variant. Third, over 500 RINRUS-designed models of the enzyme catechol-O-methyltransferase are analyzed to identify cheminformatic features that might be foundational for efficient, accurate model designs.There is the possibility to incorporate machine learning into the RINRUS workflow to enable the transformation of simple qualitative/semi-quantitative chemical characteristics into descriptors suitable for more quantitative network designs. This is illustrated in the final piece of this work where random forest models constructed from the chemical information of four proteins were able to accurately predict quantitative inter-residue interaction energies for an untested protein only using several structural, network, and chemical descriptors. Collectively, the studies illustrate the value of the RINRUS toolkit in creating practical, accurate models of enzyme active sites, and they provide direction for future improvement with the methodology.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest

Download

Share

COinS