Modeling Techniques: Computational Modeling of Drug Disposition

Computational modeling of drug disposition is a critical aspect of pharmaceutical sciences, offering a quantitative and mechanistic framework to predict how drugs are absorbed, distributed, metabolized, and excreted (ADME) within the human body. Historically, drug discovery primarily focused on efficacy and selectivity, but a significant portion of drug candidates (nearly half) failed in later clinical trials due to undesirable pharmacokinetic properties. Since the mid-1990s, there has been a paradigm shift, with a widespread adoption of early in vitro evaluation of ADMET properties to reduce attrition rates at more expensive stages. These in vitro results have enabled the training of in silico (computational) models, which can predict ADMET properties even before compounds are synthesized, offering advantages in terms of cost, time, and ethical considerations compared to traditional in vitro and in vivo experiments. Computational models translate physiological, biochemical, and molecular processes into mathematical equations to simulate drug behavior under various conditions.

There are mainly two types of modeling approaches: quantitative and qualitative. Beyond this broad categorization, several prominent and increasingly sophisticated computational techniques are employed:

Quantitative Approaches:

These methods investigate the structural requirements for the interaction between drugs and the targets involved in ADMET processes. They are especially useful when there is a significant accumulation of knowledge against a certain target.

Pharmacophore Modeling:

• Definition and Purpose:
  • A pharmacophore model identifies the essential 3D arrangements of chemical features necessary for a molecule to interact with a specific biological target, such as a receptor or a transporter protein. These features typically include hydrophobic centers, hydrogen bond donors, hydrogen bond acceptors, and positively or negatively ionizable groups, along with their geometric arrangement.
  • It represents the structural points and distance between important structural features that are critical for biological activity.
  • Pharmacophore modeling is particularly useful when there is an accumulation of knowledge against a certain target. For example, a set of drugs known to be transported by a transporter can enable a pharmacophore study to elucidate the minimum required structural features for transport.
• Integration with Quantitative Approaches:
  • Pharmacophore modeling is categorized as a quantitative approach alongside flexible docking studies. These approaches aim to investigate the structural requirements for drug-target interactions involved in ADMET processes.
  • It is a pivotal computational approach in drug discovery and for understanding biotransformation processes, often used in conjunction with molecular docking and QSAR/QSPR (Quantitative Structure-Activity/Property Relationship).
• Tools and Algorithms for Generation:
  • Three widely used automated pharmacophore perception tools are DISCO (DIStance COmparisons), GASP (Genetic Algorithm Similarity Program), and Catalyst/HIPHOP. All three programs attempt to determine common features based on the superposition of active compounds using different algorithms.
  • Catalyst system automation includes HipHop (pharmacophore model generation based on similarity index) and HypoGen (using quantitative property data of active structures).
  • DISCO emphasizes the importance of both ligand features and receptor characteristics (interactive space, heavy atoms, flexibility).
  • GASP considers molecules as single entities with randomized orientations, selecting those with the least pharmacophoric features.
  • Other software mentioned for pharmacophore model generation includes Drug Discovery studio, Ligandscout, ZINC Pharmer, and PharmaGist. PharmaGist can identify common structural features among a group of compounds and calculate bond distances.
• Applications in Drug Disposition Studies (ADMET Context):
  • Early-phase screening: Pharmacophore models are applied to large databases of drug-like molecules to identify compounds likely to interact with specific transporters or enzymes involved in ADMET processes. This helps in early assessment of pharmacokinetic-toxicity profiles.
  • Understanding Transport Requirements: They help elucidate the structural characteristics necessary for a drug to bind to or be transported by a particular protein.
  • Guiding Drug Design: By identifying key features, pharmacophore models guide the design of new drug molecules with improved ADMET properties.
• Specific Examples of Transporter Modeling:
  • P-glycoprotein (P-gp): Pharmacophore models have been generated to predict the inhibition of P-gp, identifying common chemical features such as hydrophobes, hydrogen bond acceptors, and ring aromatic features. These models aid in screening compounds for potential efflux-related bioavailability problems and identifying novel P-gp inhibitors.
  • Breast Cancer Resistance Protein (BCRP): While a 3D-QSAR model for BCRP emphasized specific structural features (e.g., 2,3-double bond in ring C, hydroxylation at position 5), the sources caution that such models, especially when based on a narrow set of structures, should be applied with care.
  • Nucleoside Transporters (CNT1, CNT2, ENT1, CNT3): Comprehensive studies generated distinctive models for these transporters, revealing common features like two hydrophobic features and one hydrogen bond acceptor on the pentose ring, as well as subtle characteristic requirements for each specific transporter.
  • hPEPT1 (Human Peptide Transporter 1): A pharmacophore model for hPEPT1 identified two hydrophobic features, one hydrogen bond donor, one hydrogen bond acceptor, and one negative ionizable feature as transport requirements. This model successfully suggested antidiabetic repaglinide and HMG-CoA reductase inhibitor fluvastatin as inhibitors.
  • ASBT (Apical Sodium-Dependent Bile Acid Transporter): A pharmacophore model revealed ASBT transport requirements as one hydrogen bond donor, one hydrogen bond acceptor, one negative charge, and three hydrophobic centers, consistent with previous 3D-QSAR studies.
  • Organic Cation Transporters (OCTs): A human OCT1 pharmacophore model suggested transport requirements as three hydrophobic features and one positive ionizable feature.
  • Organic Anion Transporting Polypeptides (OATPs): A metapharmacophore approach for OATP1B1 identified three hydrophobic features flanked by two hydrogen bond acceptor features as essential requirements.
  • BBB-Choline Transporter: Though not cloned, 3D-QSAR models for this transporter identified three hydrophobic interactions and one hydrogen bonding interaction surrounding the positively charged ammonium moiety as important for recognition.
• Challenges and Future Directions:
  • Despite advances, the currently available transporter models only cover a small fraction of all transporters involved in drug disposition.
  • Data quality remains the most limiting factor in ADMET modeling, including for pharmacophore models.
  • Incorporation of the influence of these transporters into current systemic ADMET models is an ongoing task. Some commercial programs, like GastroPlus, PK-Sim, and ADME/Tox WEB, are already implementing capabilities for modeling active transport, using pharmacophore models as filters.
  • It’s important to note that a compound fitting a transport model renders it susceptible to that transporter, but not fitting the model does not necessarily exclude it, as no model can cover all possible chemical space.

Flexible Docking Studies:

• Purpose and Definition:
  • Flexible docking is a computational process used to identify molecular interactions between a drug molecule and its receptor, which corresponds to biological activity.
  • It considers the three-dimensional movements of ligands, allowing for a more realistic simulation of how a drug fits into a dynamic binding site.
  • The goal is to affirm the pose of interactions between a small molecule and a receptor (protein or enzyme) or between two proteins.
• Relationship with Other Quantitative Approaches:
  • Flexible docking studies, along with pharmacophore modeling, constitute the primary quantitative approaches for investigating structural requirements of drug-target interactions in ADMET processes.
  • The essential interactions derived from flexible docking can be used as a screen in evaluating drug ADMET properties.
  • It is a pivotal computational approach in drug discovery and for understanding biotransformation processes, often used in conjunction with molecular modeling, molecular dynamics (MD) simulations, and QSAR/QSPR (Quantitative Structure-Activity/Property Relationship).
• Mechanism and Strategies:
  • Flexible docking considers “(6+N) special orientations” for the three-dimensional movements of ligands, often employing Monte Carlo simulation processes.
  • Four different strategies for flexible docking are mentioned: Monte Carlo molecular dynamic simulation, in-site combinatorial searching, building of ligand molecules, and site assessment with fragmentation.
  • Important factors in this process include grid space volume, relative mean standard deviation, and receptor active pocket analysis.
  • The methodology typically involves:
    1. Preparation of receptor: Receptors can be procured from sources like the Protein Data Bank (www.rcsb.org) or generated using homology modeling. Water molecules and co-crystallized ligands are removed, Gasteiger charges are added, and the file is saved in .pdbqt format.
    2. Preparation of ligand molecule: Rotation and charges are added to the ligand, saved in .pdbqt format.
    3. Grid measurement: Based on interactive residues, grid space volume, and exhaustiveness, settings are configured for iterated interactions.
    4. Drug-receptor interaction and visualization: The possible interaction between nearby amino acids and the ligand within the receptor voxel is visualized, and tabulated energy data (Kcal/mole) with relative mean standard deviation are obtained.
• Software and Tools:
  • A review of different flexible docking algorithms in drug discovery has been recently provided.
  • AutoDock Vina is highlighted as a widely used and freely available software for flexible docking.
  • Other commercial standalone software options include GLIDE, GOLD, and MedusaDock.
  • Molecular modeling and simulation tools that fall under this category also include AutoDock, and PHENIX and AFITT are mentioned for improved ligand geometries in crystallographic refinement.
• Applications in Drug Disposition Studies:
  • Molecular docking studies play a pivotal role in terms of pharmacokinetic behavior and biological activity.
  • They are used to predict drug-protein interactions, binding affinities, and conformational changes relevant to absorption, metabolism, and transport processes.
  • Specifically, molecular docking has been applied in understanding various transporters:
    • P-glycoprotein (P-gp): With the availability of P-gp crystal structures, molecular docking and molecular dynamics simulations have been used to elucidate binding modes and predict ligand affinity.
    • Nucleoside Transporters: Homology models based on related transporters are used for docking studies to predict binding and transport.
    • Organic Cation Transporters (OCTs): Structure-based models using homology modeling, molecular docking, and molecular dynamics simulations are mentioned for predicting binding affinities and transport mechanisms.
    • Organic Anion Transporting Polypeptides (OATPs): Recent cryo-EM structures of OATP1B1 and OATP1B3 enable docking studies to rationalize substrate specificity and drug-drug interactions.
    • BBB-Choline Transporter: Homology modeling based on related transporters (e.g., ChT1) is used for molecular docking studies to predict binding affinities and identify structural features important for interaction.
• Challenges:
  • Similar to other ADMET modeling, data quality (e.g., experimental binding data for training and validation) and the availability of high-resolution protein structures are critical limitations.
  • Predicting transporter-mediated drug disposition remains challenging due to limited structural data and interspecies differences.

In essence, flexible docking studies provide atomic-level insights into drug-target interactions, complementing other computational approaches like pharmacophore modeling to enhance the understanding and prediction of a drug’s disposition within the body.

• Qualitative Approaches: These approaches, such as Quantitative Structure-Activity Relationship (QSAR) and Quantitative Structure-Property Relationship (QSPR) studies, utilize multivariate analysis to correlate molecular descriptors with ADMET-related properties. A diverse range of molecular descriptors, from simple ones like molecular weight to complex quantum mechanical concepts, can be calculated based on drug structure.
  • When calculating correlations, it is crucial to select molecular descriptors that adequately represent the type of interactions contributing to the targeted biological property. The majority of published ADMET models are generated based on 2D descriptors, although alignment-dependent 3D descriptors can generate highly predictive models, but their application in high-throughput settings is challenging due to alignment difficulties.
  • Various statistical algorithms are available for correlating field descriptors with ADMET properties, including simple multiple linear regression (MLR), multivariate partial least-squares (PLS), artificial neural networks (ANN), and support vector machines (SVM). No single method consistently outperforms others, emphasizing the importance of selecting the appropriate mathematical tool. Software tools like DRAGON, E-Dragon, and PaDEL are used to calculate molecular descriptors.
• Other Significant Modeling Techniques:
  • Compartmental Models: These are classical pharmacokinetic (PK) models that simplify the body into a series of interconnected compartments (e.g., central, peripheral), with drug transfer described by rate constants. While useful for fitting experimental data, they often lack physiological interpretability. Software like Phoenix WinNonlin and PKSolver implement these models.
  • Physiologically Based Pharmacokinetic (PBPK) Models: These are mechanistic, multi-compartmental models that represent actual organs and tissues, incorporating physiological parameters like tissue volumes, blood flow rates, and enzyme/transporter expression. PBPK models can simulate drug kinetics in special populations (e.g., pediatrics, elderly, patients with organ impairment) and predict tissue-specific drug concentrations, making them invaluable for dose optimization and safety assessment. They also allow for extrapolation across species and disease states. Commercial and open-source platforms include Simcyp, GastroPlus, PK-Sim, MoBi, and Berkeley Madonna.
  • Machine Learning (ML) and Artificial Intelligence (AI): These approaches, including deep learning, random forests, and support vector machines, are increasingly applied to predict complex ADMET properties, transporter interactions, and toxicity endpoints using large, heterogeneous datasets. They can handle high-dimensional data and complex, nonlinear relationships. Key software and tools include ADMET Predictor, DeepChem, and FP-ADMET.
  • Population Pharmacokinetic (PopPK) Models: These models analyze drug concentration data from populations to quantify variability and identify factors influencing drug kinetics, which is essential for understanding inter-individual differences and optimizing dosing in diverse patient groups. Software such as NONMEM, Monolix, and Pmetrics are widely used for PopPK modeling.
  • Noncompartmental Analysis (NCA): This is a model-independent approach that estimates pharmacokinetic parameters (e.g., area under the curve, clearance, half-life) directly from concentration-time data without assuming a specific compartmental structure. It is useful for initial data analysis and is implemented in tools like Phoenix WinNonlin and PKSolver.
  • Molecular Modeling and Simulation: This category includes techniques like homology modeling, which is used when the three-dimensional structure of a target protein or receptor is not available; it develops a new protein model based on sequence similarity (e.g., using FASTA or BLAST).
  • Hybrid and Multiscale Models: These models integrate different modeling approaches (e.g., combining PBPK with ML or QSAR) to capture processes across multiple biological scales, from molecular interactions to whole-body disposition.

The development and application of these computational programs are fueled by ever-increasing computational power and significant advances in in silico modeling algorithms. The aim is to provide more accurate predictions of drug disposition behavior by incorporating the complex effects of various factors, including active transporters. Some commercial programs, such as GastroPlus and ADME/Tox WEB, have already implemented capabilities for modeling active transport.

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

pharmacyinfoline

Administrator

Visit Website View All Posts