©2020-2032 All Rights Reserved. Online Journal of Bioinformatics. You may not store these pages in any form except for your own personal
use. All other usage or distribution is illegal under international copyright
treaties. Permission
to use any of these pages in any other way besides the before mentioned must be
gained in writing from the publisher. This
article is exclusively copyrighted in its entirety to onlinejournals@gmail.com.
This article may be copied once but may not be, reproduced or re-transmitted without the express
permission of the editors. Linking: To link to this page or any pages linking
to this page you must link directly to this page only here rather than put up
your own page.
OJBTM
Online Journal of Bioinformatics©
Volume 21 (3):280-287, 2020.
A model for identifying drug like molecules
Kailash Adhikari2,
Tapobrata Lahiri1[1],
Hrishikesh Mishra1, Kalpana
Singh1, Arun Kumar CN2
1Indian
Institute of Information Technology, Jhalwa Campus,
Allahabad, 2IBM India Pvt Ltd., Bangalore-India
ABSTRACT
Adhikari K, Lahiri T, Mishra
H, Singh K, Kumar CAN., A model for identifying drug like molecules, Onl J Bioinform., 21 (3):280-287,
2020. A model to identify drug like
characteristics of any small molecule from any database is described. DRAGON software
and feature selection was used to extract 15 of 785 sets of features found to
be significant for discrimination between drug and non-drug like molecules. These
features were fed into a forward back propagation neural network classifier
whose weights and biases were optimized through a Neuro-GA module. Selection was
based on numerical values. Simple filter of 785 features generated 23% with P >
0.1, 73.8% P
< 0.05 and others 0.05 and 0.1 to yield 580 with significant discriminating
power. Classification of 600 molecules into drug and non
drug like molecules was done on 450 training and 150 as test sets. After training networks we found 88% for
training and 86% test sets. We enhanced accuracy for weight and biases by genetic
algorithm Neuro-GA, of test set to 90.7%. We confirmed efficiency with 50 new molecules
selecting the same 15 features yielding accuracy of 86% and 90%.
Key words: Drug likeness, molecular descriptors, data
warehousing and mining, backpropagation network.
FULL-TEXT (SUBSCRIBE OR
PURCHASE TITLE)