Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/156881
Title: Geometric deep learning for antibiotic discovery
Authors: Choo, Hou Yee
Keywords: Science::Medicine::Pharmacy::Pharmaceutical technology
Science::Mathematics
Issue Date: 2022
Publisher: Nanyang Technological University
Source: Choo, H. Y. (2022). Geometric deep learning for antibiotic discovery. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/156881
Abstract: Nowadays, in order to reduce the unbearable laboratory cost, time cost and increase the accuracy rate of new drug identification at the same time, Artificial Intelligence (AI) techniques have been widely applied in pharmaceutical industry for drug discovery programs. In this article, we proposed a geometric deep learning model that utilized the Graph Attention Network (GAT) to identify potential new antibiotics candidates. Then, several performance metrics were tested on to evaluate the model, which included AUC-ROC, accuracy, and weighted average of precision, recall and F1-score. The performance of the proposed model was then compared with other existing geometric deep learning models. Undersampling and 5-fold cross validation were applied to reduce imbalance of data and reduce the variance and bias of the resulting performance metrics, respectively, to make our experiment fair. The result of the experiment showed that the proposed model outperformed all other competing models in all performance metrics. This probably implies that the proposed model, which leverages more on the neighboring messages that are more relevant to the updating atoms, are more suitable for molecular property identification. Also, an ablation study was conducted to investigate the contribution of Morgan Fingerprint, molecular graph embeddings, and SMILES text embeddings towards the molecular property of interest. It turned out that Morgan Fingerprint and molecular graph embeddings are the optimal combination of embeddings to be included in our model.
URI: https://hdl.handle.net/10356/156881
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SPMS Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
fyp_final.pdf
  Restricted Access
693.31 kBAdobe PDFView/Open

Page view(s)

18
Updated on May 19, 2022

Download(s)

6
Updated on May 19, 2022

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.