Computational modeling for solubility is a critical aspect of drug absorption, as a drug must generally dissolve before it can be absorbed from the intestinal lumen. Poor solubility can lead to low and variable bioavailability, erratic pharmacokinetics, and ultimately, clinical failure. This is particularly challenging for BCS Class II (low solubility, high permeability) and Class IV (low solubility, low permeability) compounds, which represent a significant portion of new drug candidates. The Developability Classification System (DCS) further refines BCS Class II into dissolution-rate-limiting (Class-IIa) and solubility-rate-limiting (Class-IIb) compounds.
Computational models predict solubility using various approaches:
- QSPR (Quantitative Structure-Property Relationship) and QSAR (Quantitative Structure-Activity Relationship) Models:
- These are empirical approaches that utilize multivariate analysis to correlate molecular descriptors with solubility. Molecular descriptors are numerical representations of chemical structure.
- Common descriptors include molecular weight, logP, polar surface area, hydrogen bond donors/acceptors, aromaticity, and melting point. Some are closely related to a physical property and are easy to comprehend, while others are based on quantum mechanical concepts or interaction energies.
- The target property for most models is the logarithm of solubility (logS), usually in mol/L units at 25°C.
- Statistical algorithms used for correlating descriptors with solubility include simple multiple linear regression (MLR), multivariate partial least-squares (PLS), artificial neural networks (ANN), and support vector machines (SVM). No single method consistently performs better than others, sometimes requiring multiple methods for comparison to identify the best approach.
- Many models are trained and verified with databases such as AQUASOL and PhysProp.
- Mechanistic and Thermodynamic Models:
- These approaches are based on the underlying physiological processes governing solubility.
- They simulate the physical and chemical processes like dissolution, ionization, and solid-state transitions.
- The dissolution process involves the breaking up of the solute from its crystal lattice and the association of the solute with solvent molecules. Weaker interactions within the crystal lattice (lower melting point) and stronger interactions between solute and solvent molecules result in better solubility.
- The “general solubility equation” can indirectly estimate solubility by measuring a drug’s logP (log of the partition coefficient between water and n-octanol) and its melting point.
- For druglike molecules, solvent-solute interaction is a major determinant of solubility. LogP is the simplest estimation of this interaction and can be predicted by commercial programs like CLogP, which uses a fragment-based approach. Other approaches amend LogP values with additional terms to account for the contribution of solute crystal lattice energy.
- Machine Learning (ML) and Artificial Intelligence (AI) Approaches:
- These include deep learning architectures, such as graph neural networks (GNNs), which can predict solubility directly from molecular structure (e.g., SMILES notation) without requiring hand-crafted descriptors.
- ML models can capture complex, non-linear relationships and are suitable for high-throughput screening.
- Biorelevant Solubility Prediction:
- Computational models also predict solubility in biorelevant media, such as fasted state simulated intestinal fluid (FaSSIF) and fed state simulated intestinal fluid (FeSSIF). This is because solubility in these media is more predictive of in vivo absorption than solubility in pure water.
- Key Factors Influencing Solubility Modeling:
- Ionization State (pKa): The ionization state of a compound significantly affects both its solubility and permeability, thereby influencing its absorption profile. The charge of a molecule can be determined using its ionization constant value (pKa) and the environmental pH. Unionized forms are typically more lipophilic and cross membranes easily, while ionized forms generally have higher aqueous solubility. Several commercial and publicly available programs provide pKa estimation based on the input structure, including SCSpKa, Pallas/pKalc, ACD/pKa, and SPARC online calculator.
- Lipophilicity (LogP): LogP is a critical descriptor, as lipophilicity influences drug movement through cellular membranes and portal systems. ALOGP, XLOGP, MLOGP, and CLOGP are different tools for predicting LogP values, with ALOGP and CLOGP being highly acceptable.
- Software Tools:
- GastroPlus™ implements the Advanced Compartmental Absorption and Transit (ACAT) model, which calculates regional solubility along the GI tract using the Henderson–Hasselbalch equation, incorporating pKa values and the effects of bile salts and micelles.
- ADMET Predictor™ is a machine learning-based suite that predicts aqueous and biorelevant solubility, among other properties.
- SwissADME is a free web-based tool that predicts water solubility.
- Other pKa estimation tools include MoKa, CHARMM-based pKa calculation, H++Poisson-Boltzmann based pKa calculations, and PROPKA.
- Case Studies:
- R1315 (Roche): GastroPlus™ was used to model the absorption of this poorly soluble CNS drug candidate. Simulations accurately predicted that solubility had little impact on absorption and that particle size reduction would not significantly improve bioavailability, guiding formulation selection.
- Aprepitant: GastroPlus™ modeled the absorption of this BCS Class II drug, predicting a threefold increase in AUC with food, which matched clinical observations, and simulated nanoparticle formulations to reduce the food effect.
These computational models provide a comprehensive in silico prediction of a drug’s solubility, which is fundamental to understanding and predicting its oral absorption and overall pharmacokinetic behavior.
Unit 2: Computational Modeling of Drug Disposition: Podcast link
Introduction, Modeling Techniques: Drug Absorption, Solubility, Intestinal Permeation, Drug Distribution ,Drug Excretion, Active Transport; P-gp, BCRP, Nucleoside Transporters, hPEPT1, ASBT, OCT, OATP, BBB-Choline Transporter.
About the Author
