Table 3.
In-silico models and their parameters used for predicting drug penetrability
| Model | Description | Parameters involved |
|---|---|---|
| Brain Penetrability Parameters | ||
| logBB | Brain to plasma ratio (log Cbrain/log Cblood) | Correlation with quantitative structure-activity relationship data |
| logPS | BBB permeability surface area product | Correlation with quantitative structure-activity relationship data |
| logCSF | Cerebrospinal fluid to plasma ratio ((log CCSF/log Cblood) | Correlation with quantitative structure-activity relationship data |
| Molecular Descriptors | ||
| logPoct | Octanol/water partition coefficient | Hydrophobicity, H-bond donor potential |
| ΔlogP | The difference in octanol/water and cyclohexane/water partition coefficients (logPoct - logPcyc) | Low overall H-bonding ability |
| logD | Log distribution coefficient | Lipophilicity (0< logD <3) |
| Classical descriptors | Physicochemical parameters | Polar surface area; Molecular weight; Molecular size, shape, and flexibility Charge |
| P-glycoprotein substrate | High-affinity P-glycoprotein substrate probability | Efflux transport through the BBB |
| Rule-based Models | ||
| Hansch’s rule of 2 | Prediction based on octanol/water partition coefficient | Compounds having log Poct ≈2.0 have optimal brain penetration |
| Modified Lipinski’s rules for CNS penetration | Prediction based on selected molecular descriptors | H-bond donors ≤3; H-bond acceptors ≤7; molecular weight ≤400 Da; log Poct ≤5.0; 7.5< pKa <10.5 |
| CNS active drugs | Prediction based on selected molecular descriptors | Polar surface area <90 Å2 H-bond donors <3; 2.0 log Poct <5.0; molecular weight <450 Da |
| Quantitative Structure-Activity Relationship (QSAR) | ||
| Linear QSAR | Prediction based on selected molecular descriptors | Multiple Linear Regression (MLR); Partial Least-Squares (PLS) methods; Variable Selection and Modelling Method based on the Prediction (VSMP); Linear Discriminant Analysis (LDA); Comprehensive Descriptors for Structural and Statistical; Analysis (CODESSA) |
| Non-linear QSAR | Prediction based on selected molecular descriptors | Neural Networks (NN); Bayesian Modelling; Support Vector Machine (SVM); Gaussian Processes; k Nearest Neighbour Method; Recursive Partitioning; Substructure Analysis |
Abbreviations: H, bond-hydrogen bond; logBB, brain to plasma ratio; logCSF, cerebrospinal fluid to plasma ratio; logD, log distribution coefficient; logP, log octanol/water partition coefficient; logPS, Blood-Brain-Barrier permeability surface area product; pKa, log of acidic dissociation constant.