Application of Artificial Neural Networks to Predict the Intrinsic Solubility of Drug-Like Molecules

Abstract

Machine learning (ML) approaches are receiving increasing attention from pharmaceutical companies and regulatory agencies, given their ability to mine knowledge from available data. In drug discovery, for example, they are employed in quantitative structure–property relationship (QSPR) models to predict biological properties from the chemical structure of a drug molecule. In this paper, following the Second Solubility Challenge (SC-2), a QSPR model based on artificial neural networks (ANNs) was built to predict the intrinsic solubility (logS0) of the 100-compound low-variance tight set and the 32-compound high-variance loose set provided by SC-2 as test datasets. First, a training dataset of 270 drug-like molecules with logS0 value experimentally determined was gathered from the literature. Then, a standard three-layer feed-forward neural network was defined by using 10 ChemGPS physico-chemical descriptors as input features. The developed ANN showed adequate predictive performances on both of the SC-2 test datasets. Benefits and limitations of ML approaches have been highlighted and discussed, starting from this case-study. The main findings confirmed that ML approaches are an attractive and promising tool to predict logS0; however, many aspects, such as data quality, molecular descriptor computation and selection, and assessment of applicability domain, are crucial but often neglected, and should be carefully considered to improve predictions based on ML.

Intrinsic Aqueous Solubility: Mechanistically Transparent Data-Driven Modeling of Drug Substances

“…Tight and loose test sets from SC2019 have been independently used four times to test new and existing models [ 20 , 27 , 28 , 29 ]. Avdeef [ 20 ] has collected aqueous solubility data from 1325 literature sources (multi-source compilations, single-source measurements for many compounds, and miscellaneous primary sources) and curated and analyzed this data to obtain 6355 intrinsic aqueous solubility values for 3014 different chemical compounds.…”

“…They found that despite their training set having inconsistencies related to pH, solid form, and temperature, the developed model had comparable prediction capability to the top-ranked models in SC2019. Tosca et al [ 28 ] developed an artificial neural networks QSPR model for intrinsic aqueous solubility and predicted both test sets with a sufficient prediction accuracy. Francoeur and Koes [ 29 ] used molecule attention transformer (MAT) architecture to develop a model (called SolTranNet) for data from AqSolDB.…”

Polymeric nanoparticles—Promising carriers for cancer therapy

“…It has been noticed for long that clinical application of many potent remedial drugs in cancer and other diseases have been limited due to the following reasons: 1) poor aqueous solubility, and consequently, minimal systemic bioavailability (Anitha et al, 2011;Tosca et al, 2021). During the treatment, patients need to be injected with a large amount of saline, which easily causes water intoxication (Lauersen and Birnbaum, 1975).…”

Artificial Neural Networks to Predict the Apparent Degree of Supersaturation in Supersaturated Lipid-Based Formulations: A Pilot Study

“…Over the last decade, interest regarding the use of ML algorithms across diverse disciplines in pharmaceutical design and development has grown [ 11 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 ]. While ML models have been produced to optimise lipid-based formulation (LBF) development [ 3 , 22 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ], the application of more novel ML approaches for bio-enabling formulations currently focuses on solid dispersions (SDs) [ 21 , 34 , 35 ].…”