Automated Framework for Developing Predictive Machine Learning Models for Data-Driven Drug Discovery

Bruno Neves,Joyce Borba,Carolina Andrade,Arthur Silva,José Moreira-Filho,Rodolpho Braga,Vinicius Alves,Melina Mottin,Eugene Muratov

doi:10.21577/0103-5053.20200160

Abstract

The increasing availability of extensive collections of chemical compounds associated with experimental data provides an opportunity to build predictive quantitative structure-activity relationship (QSAR) models using machine learning (ML) algorithms. These models can promote data-driven decisions and have the potential to speed up the drug discovery process and reduce their failure rates. However, many essential aspects of data preparation and modeling are not available in any standalone program. Here, we developed an automated framework for the curation of chemogenomics data and to develop QSAR models for virtual screening using the open-source KoNstanz Information MinEr (KNIME) program. The workflow includes four modules: (i) dataset preparation and curation; (ii) chemical space analysis and structure-activity relationships (SAR) rules; (iii) modeling; and (iv) virtual screening (VS). As case studies, we applied these workflows to four datasets associated with different endpoints. The implemented protocol can efficiently curate chemical and biological data in public databases and generates robust QSAR models. We provide scientists a simple and guided cheminformatics workbench following the best practices widely accepted by the community, in which scientists can adapt to solve their research problems. The workflows are freely available for download at GitHub and LabMol web portals.

Highlights

Quantitative Structure-Activity Relationship (QSAR) modeling is a major cheminformatics approach in computer-aided drug discovery.[1,2] Nowadays, machine learning (ML) methods can be used to generate QSAR models that accurately predict chemicals and how chemical modifications might influence biological properties.[2]
We provide scientists a simple and guided cheminformatics workbench following the best practices widely accepted by the community, in which scientists can adapt to solve their research problems
We developed an automated computational workflow for building robust and predictive QSAR models employing ML algorithms following the best practices for model development and validation

Summary

Introduction

Quantitative Structure-Activity Relationship (QSAR) modeling is a major cheminformatics approach in computer-aided drug discovery.[1,2] Nowadays, machine learning (ML) methods can be used to generate QSAR models that accurately predict chemicals and how chemical modifications might influence biological properties.[2]. The development and implementation of high-quality models require that users have a thorough understanding of the modeled bioassay data, expert comprehension of best practices for model development, validation and application,[21] and computational skills. Until this date, many essential aspects of data preparation and modeling are not available in a single standalone software. We developed an automated computational framework to curate and prepare datasets, to generate and validate predictive ML models, and to perform virtual screening (VS) of chemical libraries using the KoNstanz Information MinEr (KNIME) software. 22,23 KNIME is an open-source platform that provides a customizable framework for data management and modeling through a user-friendly graphical interface. 24,25

Results and Discussion

Literature

Conclusions