An automated framework for QSAR model building

Abstract

BackgroundIn-silico quantitative structure–activity relationship (QSAR) models based tools are widely used to screen huge databases of compounds in order to determine the biological properties of chemical molecules based on their chemical structure. With the passage of time, the exponentially growing amount of synthesized and known chemicals data demands computationally efficient automated QSAR modeling tools, available to researchers that may lack extensive knowledge of machine learning modeling. Thus, a fully automated and advanced modeling platform can be an important addition to the QSAR community.ResultsIn the presented workflow the process from data preparation to model building and validation has been completely automated. The most critical modeling tasks (data curation, data set characteristics evaluation, variable selection and validation) that largely influence the performance of QSAR models were focused. It is also included the ability to quickly evaluate the feasibility of a given data set to be modeled. The developed framework is tested on data sets of thirty different problems. The best-optimized feature selection methodology in the developed workflow is able to remove 62–99% of all redundant data. On average, about 19% of the prediction error was reduced by using feature selection producing an increase of 49% in the percentage of variance explained (PVE) compared to models without feature selection. Selecting only the models with a modelability score above 0.6, average PVE scores were 0.71. A strong correlation was verified between the modelability scores and the PVE of the models produced with variable selection.ConclusionsWe developed an extendable and highly customizable fully automated QSAR modeling framework. This designed workflow does not require any advanced parameterization nor depends on users decisions or expertise in machine learning/programming. With just a given target or problem, the workflow follows an unbiased standard protocol to develop reliable QSAR models by directly accessing online manually curated databases or by using private data sets. The other distinctive features of the workflow include prior estimation of data modelability to avoid time-consuming modeling trials for non modelable data sets, an efficient variable selection procedure and the facility of output availability at each modeling task for the diverse application and reproduction of historical predictions. The results reached on a selection of thirty QSAR problems suggest that the approach is capable of building reliable models even for challenging problems.Electronic supplementary materialThe online version of this article (10.1186/s13321-017-0256-5) contains supplementary material, which is available to authorized users.

Towards reproducible computational drug discovery

“…Nantasenamat et al [257] reported the development of an automated data mining software for QSAR modeling called AutoWeka that is based on the machine learning software Weka [258]. Kausar and Falcao [259] presents an automated framework based on KNIME for QSAR modeling entailing data pre-processing, model building and validation. Dong et al [260] introduced an online platform for QSAR modeling known as ChemSAR that is capable of handling chemical structures, computing molecular descriptors, model building as well as producing result plots.…”

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

“…In addition, we tested the workflow by developing four case studies for the prediction of antiplasmodial and antischistosomal activity, as well as cardiotoxicity and mutagenicity. Previously, other groups have proposed automated 26 and semi-automated 27 KNIME workflows for the development of ML models and cheminformatics analysis. Similar workflows 28 using the commercial Pipeline Pilot 29 software have also been published.…”

Virtual Screening of DPP-4 Inhibitors Using QSAR-Based Artificial Intelligence and Molecular Docking of Hit Compounds to DPP-8 and DPP-9 Enzymes

“…The Dataset containing 4,661 compounds were curated using the nM (nanomolar) activity unit by removing the empty activity value, salt, small fragments. The molecular structures were then normalized, followed by final checking for compound duplication [14,37]. The remaining 3,933 compounds were then used for the regression modeling dataset.…”

“…For the QSAR classification models 4,355 compounds were used from the ChEMBL database with a scientific literature filter. The missing value and duplication of the dataset were curated, the salt and small fragments were removed, the molecular structure was normalized, and the duplication was removed [14,37], leaving 3,740 compounds. Furthermore, the data are classified into active and inactive compounds that used for modeling, with pIC50 activity above 7.5 was active compounds, compounds with pIC50 below 6 were inactive compounds, and compounds with pIC50 between 7.5 and 6 were grey compounds that removed [28].…”

“…In this study, virtual screening of the database will be done by building a virtual screening workflow [14][15][16][17][18], which uses the QSAR classification models to predict active compounds from the database and then uses the QSAR regression method to see the value of its activity [16]. The progress of QSAR methods can benefit from modern artificial intelligence (AI) approaches to develop computation models [19][20][21][22], this study focuses on the development of a QSAR method based on AI with a supervised learning algorithm.…”