Dissecting Machine-Learning Prediction of Molecular Activity: Is an Applicability Domain Needed for Quantitative Structure–Activity Relationship Models Based on Deep Neural Networks?

2018 ◽  
Vol 59 (1) ◽  
pp. 117-126 ◽  
Author(s):  
Ruifeng Liu ◽  
Hao Wang ◽  
Kyle P. Glover ◽  
Michael G. Feasel ◽  
Anders Wallqvist
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Deep Neural Networks ◽  
Quantitative Structure Activity Relationship ◽  
Structure Activity Relationship ◽  
Applicability Domain ◽  
Activity Relationship ◽  
Quantitative Structure ◽  
Structure Activity

scholarly journals Quantitative structure–activity relationship to predict the anti-malarial activity in a set of new imidazolopiperazines based on artificial neural networks

2019 ◽  
Vol 18 (1) ◽  
Author(s):  
Saeed Yousefinejad ◽  
Marjan Mahboubifar ◽  
Rayhaneh Eskandari
Keyword(s):  
Neural Networks ◽  
Artificial Neural Networks ◽  
Structural Information ◽  
Quantitative Structure Activity Relationship ◽  
Structure Activity Relationship ◽  
Activity Relationship ◽  
Quantitative Structure ◽  
Structure Activity ◽  
Marquardt Algorithm ◽  
Artificial Neural

Abstract Background After years of efforts on the control of malaria, it remains as a most deadly infectious disease. A major problem for the available anti-malarial drugs is the occurrence of drug resistance in Plasmodium. Developing of new compounds or modification of existing anti-malarial drugs is an effective approach to face this challenge. Quantitative structure activity relationship (QSAR) modelling plays an important role in design and modification of anti-malarial compounds by estimation of the activity of the compounds. Methods In this research, the QSAR study was done on anti-malarial activity of 33 imidazolopiperazine compounds based on artificial neural networks (ANN). The structural descriptors of imidazolopiperazine molecules was used as the independents variables and their activity against 3D7 and W2 strains was used as the dependent variables. During modelling process, 70% of compound was used as the training and two 15% of imidazolopiperazines were used as the validation and external test sets. In this work, stepwise multiple linear regression was applied as the valuable selection and ANN with Levenberg–Marquardt algorithm was utilized as an efficient non-linear approach to correlate between structural information of molecules and their anti-malarial activity. Results The sufficiency of the suggested method to estimate the anti-malarial activity of imidazolopiperazine compounds at two 3D7 and W2 strains was demonstrated using statistical parameters, such as correlation coefficient (R2), mean square error (MSE). For instance R2train = 0.947, R2val = 0.959, R2test = 0.920 shows the potential of the suggested model for the prediction of 3D7 activity. Different statistical approaches such as and applicability domain (AD) and y-scrambling was also showed the validity of models. Conclusion QSAR can be an efficient way to virtual screening the molecules to design more efficient compounds with activity against malaria (3D7 and W2 strains). Imidazolopiperazines can be good candidates and change in the structure and functional groups can be done intelligently using QSAR approach to rich more efficient compounds with decreasing trial–error runs during synthesis.

Sign in / Sign up

Export Citation Format

Share Document