ADME Prediction with KNIME: In silico aqueous solubility models based on supervised recursive machine learning approaches

Abstract

In-silico prediction of aqueous solubility plays an important role during the drug discovery and development processes. For many years, the limited performance of in-silico solubility models has been attributed to the lack of high-quality solubility data for pharmaceutical molecules. However, some studies suggest that the poor accuracy of solubility prediction is not related to the quality of the experimental data and that more precise methodologies (algorithms and/or set of descriptors) are required for predicting aqueous solubility for pharmaceutical molecules. In this study a large and diverse database was generated with aqueous solubility values collected from two public sources; two new recursive machine-learning approaches were developed for data cleaning and variable selection, and a consensus model based on regression and classification algorithms was created. The modeling protocol, which includes the curation of chemical and experimental data, was implemented in KNIME, with the aim of obtaining an automated workflow for the prediction of new databases. Finally, we compared several methods or models available in the literature with our consensus model, showing results comparable or even outperforming previous published models.

ADME prediction with KNIME: A retrospective contribution to the second “Solubility Challenge”

“…The parameters of all models were optimized based on the RMSE minimization of the RELIABLE test set. Full details on our developed algorithm are given in previous published paper [ 4 ].…”

“…The InChi (IUPAC International Chemical Identifier) code was used for duplicate identification and the standard deviation among experimental measurements was computed. A detailed description of this procedure has been shown in our previous article [4]. Although the hypothesis that -the A d v a n c e d o n l i n e a r t i c l e quality of the experimental data is the main limiting factor in predicting aqueous solubility-has been challenged [12], any variability in the experimental protocol is always "noise" for in silico modelling purposes.…”

“…In our previous publication, we presented a new method based on recursive random forest approaches to predict aqueous solubility values of drug and drug-like molecules [ 4 ]. It was based on the development of two novel recursive machine-learning approaches used for data cleaning and variable selection, and a consensus model generated by the combination of regression and classification algorithms.…”

“…The present study describes the performance of our model with the molecules of the second “Solubility Challenge” and the comparison of the results with those obtained with the best performing models of the competition. It is necessary to clarify that, for this task, the model was not trained, retrained or optimized based on the molecules of the challenge tests, i.e., the model parameters or hyper-parameters remained exactly the same as those set in previously published work [ 4 ].…”

VLVS: A Knime Based Virtual Library Generation and Screening Workflow

“…Thanks to its capabilities that can be expanded with add-ons, it has a wide potential for use in life sciences. [24] Next generation sequencing [25], metabolomics analysis [26], QSPR [27], QSAR [28], high content screening [29] and drug discovery studies [30][31][32] are examples of wide range of uses.…”