Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/156576
Title: Predicting drug responses from extended-connectivity fingerprints (ECFPs) of drugs by using graph neural networks
Authors: Asok Kumar Gaurav
Keywords: Engineering::Computer science and engineering
Issue Date: 2022
Publisher: Nanyang Technological University
Source: Asok Kumar Gaurav (2022). Predicting drug responses from extended-connectivity fingerprints (ECFPs) of drugs by using graph neural networks. Final Year Project (FYP), Nanyang Technological University, Singapore. https://hdl.handle.net/10356/156576
Project: SCSE21-0209
Abstract: Cancer is one of the leading causes of deaths and one of the treatments of cancer cure is the use of drugs that inhibits the growth of cancer tumors. Hance, it is essential to be able to predict the response of cancer drugs to assess the effectiveness of each drug in slowing down the growth of abnormal malignant cancer cells. Using computational models to predicting the response of each cancer drug can improve the chances of the patient’s successful recovery Currently, there is no ideal cancer drug response prediction method. Hence, in this report we are predicting the drug responses using drug the Extended-Connectivity Fingerprints (ECFPs) of cancer drugs as input to a graph neural network. Extended-Connectivity Fingerprints (ECFPs) of a drug captures the substructural information the drug, and hence gives a numerical representation any drug from its molecular formula. In previous research, to model the molecular graphs with GNN, certain atom features, such as atom symbol, atom degree, etc., of drugs were applied to form the node features of GNN. In this project, Extended-Connectivity Fingerprints (ECFPs) are also extracted to extend the node features. We demonstrate that GNN using ECFPs features of drugs are better in representing drug chemical features than using simple atom features alone. In this project, three types of GNN are implemented for prediction of drug responses from chemical features: Graph Convolutional Networks (GCN), GraphSAGE, and Graph Attention Networks (GAT).
URI: https://hdl.handle.net/10356/156576
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
FYP_Report_AsokKumarGaurav_U1823850F[SCSE21-0209].pdf
  Restricted Access
1.76 MBAdobe PDFView/Open

Page view(s)

15
Updated on May 20, 2022

Download(s)

6
Updated on May 20, 2022

Google ScholarTM

Check

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