Open AccessQuantitative Structure-Activity Relationship Modeling Based on Improving Penalized Linear Regression Model
Author(s) -
Rehad Shamany,
Zakariya Yahya Algamal
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1897/1/012016
Subject(s) - quantitative structure–activity relationship , applicability domain , computer science , curse of dimensionality , model selection , artificial intelligence , machine learning , linear model , linear regression , feature selection , data mining , mathematics
One of the powerful and a promising model which is used to understand the structural relationship between the chemical activity and the chemical compounds is the quantitative structure-activity relationship (QSAR). However, the huge in dimensionality is one of the major problems which affect the quality of the QSAR modeling. Penalized methods are an attractive framework that have been adapted and gained popularity among researchers as the key for performing descriptor selection and QSAR model estimation simultaneously. The choice of the tuning parameter of the penalized methods is critical. Our aim of this paper is to efficiently estimate such a tuning parameter by using bat algorithm (BA), which is a king of nature-inspired algorithms. Experimental results, obtained by running on two datasets, show that our proposed method performs better than other methods, in terms of prediction, number of selected descriptors, and running time. Further, the Y-randomization test and applicability domain confirm that the constructed QSAR model by BA method is reliable and robust.