Prediction of aqueous intrinsic solubility of druglike molecules using Random Forest regression trained with Wiki-pS0 database

Avdeef, Alex

doi:10.5599/admet.766

Cited by 45 publications

(138 citation statements)

References 158 publications

(181 reference statements)

Supporting

Mentioning

125

Contrasting

Order By: Relevance

“…This is an important characteristic differentiating the two groups. In the small-molecule set, the NHA and NHD groups overlap considerably, as illustrated elsewhere [20]. But, in the big-molecule set (Fig.…”

Section: Physicochemical Properties Of the Big Moleculesmentioning

confidence: 81%

“…Of the new machine-learning statistical approaches, the Random Forest regression (RFR) method is thought to be one of the most accurate in predicting solubility [17][18][19][20]. RFR can be employed 'off the shelf,' requiring only minimal learning [19].…”

Section: Random Forest Regressionmentioning

confidence: 99%

“…Many methods for predicting solubility have been described [5][6][7], based on linear quantitative structure-property relationship (QSPR) approaches [8][9][10][11][12][13][14][15][16]. More recent methods have evolved using machine learning statistical methods [17][18][19][20]. The molecular descriptors needed for these predictions are most often calculated from two-dimensional (2D) structures.…”

Section: Introductionmentioning

confidence: 99%

“…Big-molecule property distributions: (a) clogP, (b) molecular weight (M W ), and (c) number of H-bond donors (NHD) and acceptors (NHA). The separation between the groups is greater than that found in small molecules[20].GSE applied to big compoundsHydrophilicity (clogP) effect in big lipophilic moleculesThe linear dependence of log S 0 on clogP(Fig. 2)was further analyzed in the context of Eq.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Can small drugs predict the intrinsic aqueous solubility of ‘beyond Rule of 5’ big drugs?

Avdeef¹,

Kansy²

2020

ADMET DMPK

Self Cite

View full text Add to dashboard Cite

The aim of the study was to explore to what extent small molecules (mostly from the Rule of 5 chemical space) can be used to predict the intrinsic aqueous solubility, S0, of big molecules from beyond the Rule of 5 (bRo5) space. It was demonstrated that the General Solubility Equation (GSE) and the Abraham Solvation Equation (ABSOLV) underpredict solubility in systematic but slightly ways. The Random Forest regression (RFR) method predicts solubility more accurately, albeit in the manner of a ‘black box.’ It was discovered that the GSE improves considerably in the case of big molecules when the coefficient of the log P term (octanol-water partition coefficient) in the equation is set to -0.4 instead of the traditional -1 value. The traditional GSE underpredicts solubility for molecules with experimental S0 < 50 µM. In contrast, the ABSOLV equation (trained with small molecules) underpredicts the solubility of big molecules in all cases tested. It was found that the errors in the ABSOLV-predicted solubilities of big molecules correlate linearly with the number of rotatable bonds, which suggests that flexibility may be an important factor in differentiating solubility of small from big molecules. Notably, most of the 31 big molecules considered have negative enthalpy of solution: these big molecules become less soluble with increasing temperature, which is compatible with ‘molecular chameleon’ behavior associated with intramolecular hydrogen bonding. The X‑ray structures of many of these molecules reveal void spaces in their crystal lattices large enough to accommodate many water molecules when such solids are in contact with aqueous media. The water sorbed into crystals suspended in aqueous solution may enhance solubility by way of intra-lattice solute-water interactions involving the numerous H‑bond acceptors in the big molecules studied. A ‘Solubility Enhancement–Big Molecules’ index was defined, which embodies many of the above findings.

show abstract

Section: Physicochemical Properties Of the Big Moleculesmentioning

confidence: 81%

Section: Random Forest Regressionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Can small drugs predict the intrinsic aqueous solubility of ‘beyond Rule of 5’ big drugs?

Avdeef¹,

Kansy²

2020

ADMET DMPK

Self Cite

View full text Add to dashboard Cite

show abstract

“…Performing high-quality solubility measurements is a difficult task due to uncertainties in experimental procedures, as explained in detail in Ref. [18]. Additionally, unintentional misprints, such as the erroneous conversions of values or units while carrying them from one source to another, cause deterioration in the quality of data.…”

Section: The Quality Of Datamentioning

confidence: 99%

Pushing the limits of solubility prediction via quality-oriented data selection

Sorkun

Koelman

2020

Preprint

View full text Add to dashboard Cite

Accurate prediction of the solubility of chemical substances in solvents remains a challenge. The sparsity of high-quality solubility data is recognized as the biggest hurdle in the development of robust data-driven methods for practical use. Nonetheless, the effects of the quality and quantity of data on aqueous solubility predictions have not yet been scrutinized. In this study, the roles of the size and the quality of datasets on the performances of the solubility prediction models are unraveled, and the concepts of actual and observed performances are introduced. In an effort to curtail the gap between actual and observed performances, a quality-oriented data selection method, which evaluates the quality of data and extracts the most accurate part of it through statistical validation, is designed. Applying this method on the largest publicly available solubility database and using a consensus machine learning approach, a top-performing solubility prediction model is achieved.

show abstract

Machine learning in prediction of intrinsic aqueous solubility of drug‐like compounds: Generalization, complexity, or predictive ability?

Lovrić

Pavlović

Žuvela

et al. 2021

Journal of Chemometrics

View full text Add to dashboard Cite

We present a collection of publicly available intrinsic aqueous solubility data of 829 drug-like compounds. Four different machine learning algorithms (random forests [RF], LightGBM, partial least squares, and least absolute shrinkage and selection operator [LASSO]) coupled with multistage permutation importance for feature selection and Bayesian hyperparameter optimization were used for the prediction of solubility based on chemical structural information. Our results show that LASSO yielded the best predictive ability on an external test set with a root mean square error (RMSE) (test) of 0.70 log points, an R 2 (test) of 0.80, and 105 features. Taking into account the number of descriptors as well, an RF model achieves the best balance between complexity and predictive ability with an RMSE(test) of 0.72 log points, an R 2 (test) of 0.78, and with only 17 features. On a more aggressive test set (principal component analysis [PCA]-based split), better generalization was observed for the RFmodel. We propose a ranking score for choosing the best model, as test set performance is only one of the factors in creating an applicable model. The ranking score is a weighted combination of generalization, number of features, and test performance. Out of the two best learners, a consensus model was built exhibiting the best predictive ability and generalization with RMSE(test) of 0.67 log points and a R 2 (test) of 0.81.

show abstract

Prediction of aqueous intrinsic solubility of druglike molecules using Random Forest regression trained with Wiki-pS0 database

Cited by 45 publications

References 158 publications

Can small drugs predict the intrinsic aqueous solubility of ‘beyond Rule of 5’ big drugs?

Can small drugs predict the intrinsic aqueous solubility of ‘beyond Rule of 5’ big drugs?

Pushing the limits of solubility prediction via quality-oriented data selection

Machine learning in prediction of intrinsic aqueous solubility of drug‐like compounds: Generalization, complexity, or predictive ability?

Contact Info

Product

Resources

About