|
Design of a bioinformatics model to predict drug compound properties and its application in inhibition of HIV replication and BACE-1
|
Karim Abbasi   , Ali Masoudi-Nejad * |
| Laboratory of system Biology and Bioinformatics (LBB), Institute of Biochemistry and Biophysics, University of Tehran, Iran , amasoudin@ut.ac.ir |
|
|
Abstract: (3355 Views) |
| In this paper, a new method for the problem of predicting the compound molecule properties in the lead optimization step in drug design is presented. In the lead optimization step, the amount of available biological data on small molecule compounds is low. In recent years, this challenge has been considered and transfer learning and deep learning techniques have been used to solve it. For this purpose, similar data sets have been used as auxiliary data to learn a reliable model. In this method, compound feature extraction plays an essential role in transferring knowledge from similar (auxiliary) data sets to the target data set. In this paper, the effect of using Edge weighted Graph Convolutional Network (EGCN) is assessed which able to consider the feature vector of the compound bond as well as the atom feature vector. To evaluate the method, we have applied the proposed approach on BACE and HIV datasets. The obtained results show that the proposed method is able to extract more efficient knowledge from similar data sets to transfer to the target data set. |
|
| Keywords: Edge weighted Graph Convolutional Network, Molecular bonds, Deep learning, Transfer learning. |
|
|
Full-Text [PDF 861 kb]
(1308 Downloads)
|
Type of Study: Research |
Subject:
Divers
Received: 2021/01/24 | Accepted: 2021/02/28 | Published: 2021/03/1
|
|
|
|
|
|
|
| Add your comments about this article |
|
|
|
