Please login first

Intelligent consensus predictor: Towards more precise predictions for external set compounds
Kunal Roy 1 , Pravin Ambure 1 , Supratik Kar 2 , Probir Ojha 1

1  Drug Theoretics and Cheminformatics Laboratory, Department of Pharmaceutical Technology, Jadavpur University, Kolkata 700 032, India
2  Interdisciplinary Center for Nanotoxicity, Department of Chemistry, Physics and Atmospheric Sciences, Jackson State University, Jackson, MS-39217, USA

Published: 11 October 2017 by MDPI AG in MOL2NET 2017, International Conference on Multidisciplinary Sciences, 3rd edition in MOL2NET 2017, International Conference on Multidisciplinary Sciences, 3rd edition
MDPI AG, 10.3390/mol2net-03-04619
Abstract:

Quantitative structure-activity relationship (QSAR) modeling has travelled a long journey in drug discovery process as well as in prediction of property and/or toxicity data of diverse chemicals in order to fill the data gaps. The goodness-of-fit and quality of a  model and its prediction capability for untested compounds are assessed through diverse validation metrics. There is a constant endeavor among QSAR researchers to get better the quality of predictions for lowering the predicted residuals for external compounds. The objective of the present study has been to improve the prediction quality for external compounds with implication of “intelligent” consensus modeling approach. Three different forms of consensus models were developed for six different datasets to explore their prediction capability on query chemicals. The types are average of predictions from all qualifying individual models (CM1), weighted average predictions from all qualifying individual models (CM2), and best selection of predictions (compound-wise) from individual models (CM3). Among three consensus models, newer strategies like CM2 and CM3 are evolved as the “winners” considering prediction errors of query compounds for the studied six data sets irrespective of diverse responses, number of data points as well as dissimilar modeling algorithm. We have also developed a tool named “Intelligent Consensus Predictor” which is freely accessible via the web http://teqip.jdvu.ac.in/QSAR_Tools/ and http://dtclab.webs.com/software-tools. The details of this work have been presented in Conferentia Chemometrica http://cc2017.ttk.mta.hu/ in Hungary during September 3-6, 2017.

References

  1. Dearden JC. The history and development of quantitative structure-activity relationships (QSARs). IJQSPR. 2016;1(1):1–44.
  2. Roy K, Kar S, Das RN. Understanding the basics of QSAR for applications in pharmaceutical sciences and risk assessment. Academic press. 2015.
  3. http://teqip.jdvu.ac.in/QSAR_Tools/
  4. http://dtclab.webs.com/software-tools
  5. Roy K, Das RN, Ambure P, Aher RB. Be aware of error measures. Further studies on validation of predictive QSAR models. Chemom Intell Lab Syst. 2016;152:18-33.
  6. Roy K, Ambure P, Kar S, Ojha PK, Is it possible to improve the quality of predictions from an intelligent” use of multiple QSAR/QSPR/QSTR models? J Chemom 2017 (Submitted)

Comments on this paper Get comment updates
Currently there are no comments available.