Wang, Luxuan
(2024)
Development and Testing of Advanced Drug Discovery Tools in the Era of AI.
Master's Thesis, University of Pittsburgh.
(Unpublished)
Abstract
Drug discovery and development is a time-consuming and costly process, typically taking 10-15 years and involving millions to billions of dollars for a new drug to finally reach the market. However, the advent of computer-aided drug design (CADD) in the latter part of the 20th century introduced a transformative force in this field. It utilizes computer and mathematical models to simulate and predict the interactions between drugs and biological molecules, thereby expediting the process of drug discovery and design. Nevertheless, challenges concerning precision and computational cost remain. During the 21st century, the rapid rise and advancement of artificial intelligence (AI)-based technology has presented a promising solution, effectively enhancing and complementing CADD. AI's ability to emulate human brain learning processes through sophisticated algorithms and extensive datasets holds significant potential for enhancing prediction accuracy and expediting drug discovery efforts. To underscore the remarkable potential of AI in this domain, we developed and validated two advanced drug discovery tools, harnessing the power of AI. The first tool focuses on addressing a critical challenge in drug discovery: the identification of colloid aggregators. This project involves developing a robust and generalizable machine learning classifier for the identification of promiscuous aggregating inhibitors. Additionally, an innovative approach named Global Sensitivity Analysis (GSA) has been introduced for interpreting model prediction results. Through GSA analysis, important substructures and physicochemical properties associated with aggregator formation are discerned, furnishing invaluable insights for the hit screening process. Another one is dedicated to enhancing docking performance, as the precise prediction of binding affinity and mode remains problematic for existing docking programs. In this investigation, a novel docking pipeline that merges our unique geometry optimization algorithm, the conjugate gradient with backtracking line search (CG-BS), with highly precise machine-learning potential ANI-2x has been innovated. It serves to further optimize and reranking the binding poses predicted by Glide software. The promising results show that ANI-2x/CG-BS holds considerable potential for integration into virtual screening pipelines due to its enhanced docking performance. These innovations exemplify the pivotal role of AI-based technologies in propelling drug discovery and development into a new era of effectiveness and efficiency.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
25 April 2024 |
| Date Type: |
Publication |
| Defense Date: |
9 April 2024 |
| Approval Date: |
25 April 2024 |
| Submission Date: |
18 April 2024 |
| Access Restriction: |
2 year -- Restrict access to University of Pittsburgh for a period of 2 years. |
| Number of Pages: |
73 |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Pharmacy > Pharmaceutical Sciences |
| Degree: |
MS - Master of Science |
| Thesis Type: |
Master's Thesis |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
Computer-aided drug design, Artificial intelligence (AI)-based technology, Machine learning, Drug discovery |
| Date Deposited: |
25 Apr 2024 16:44 |
| Last Modified: |
25 Apr 2024 16:44 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/46165 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[img]](https://d-scholarship.pitt.edu/style/images/fileicons/application_pdf.png)
![[feed]](/images/feed-icon-32x32.png){kind=icon}