TL;DR
This paper presents a novel hybrid virtual screening approach combining machine learning and structure-based methods to discover new inhibitors of butyrylcholinesterase (BChE), a key target for Alzheimer's disease (AD) treatment. The study successfully identified two promising drug candidates, piboserod and Rotigotine, demonstrating potent BChE inhibition, neuroprotective effects, and favorable safety profiles, with Rotigotine being a repurposed existing drug.
What Problem Does This Paper Address?
Alzheimer's disease (AD) is a progressive neurodegenerative disorder characterized by cognitive decline, and current treatments offer only symptomatic relief and often come with side effects. Targeting the cholinergic system, specifically by inhibiting enzymes that degrade acetylcholine, is a promising therapeutic strategy. While acetylcholinesterase (AChE) has been the primary target, butyrylcholinesterase (BChE) is gaining attention due to its increased activity in AD patients and its potential role in disease progression. However, traditional drug discovery methods are time-consuming and expensive. This paper addresses the need for efficient and accurate methods to identify novel BChE inhibitors, particularly those with favorable drug-like properties and potential for clinical translation.
Methods
The researchers employed a multi-pronged approach:
- Machine Learning-based QSAR Model: They developed a quantitative structure-activity relationship (QSAR) model using various machine learning algorithms (XGBoost, RF, SVM, KNN) to predict BChE inhibitory activity. The model was trained on a dataset of known BChE inhibitors and non-inhibitors, using ECFP4 molecular fingerprints. XGBoost emerged as the best performer.
- Hybrid Virtual Screening: This involved integrating:
- Ligand-Based Virtual Screening (LBVS): Using the developed QSAR model to filter a large compound library.
- Structure-Based Virtual Screening (SBVS): Employing molecular docking to assess the binding affinity of filtered compounds to the BChE protein.
- In Silico ADME and BBB Prediction: To prioritize drug-like candidates, they predicted absorption, distribution, metabolism, excretion (ADME) properties and blood-brain barrier (BBB) permeability.
- In Vitro Biological Evaluation: Promising hits were experimentally tested for BChE inhibitory potency (IC50), cytotoxicity, and neuroprotective effects in PC-12 cells.
- Enzyme Kinetics Analysis: Detailed kinetic studies were performed to understand the mechanism of inhibition for the top compounds.
- Spectroscopic and Molecular Dynamics Studies: Fluorescence quenching, 3D fluorescence spectroscopy, and molecular dynamics simulations were used to elucidate the binding modes and stability of the inhibitors with BChE.
Datasets
- Training Dataset: Derived from the ChEMBL database, containing molecular information and activity data for human-derived BChE enzyme. Compounds with IC50 < 100 nM were labeled as active (1), and those with IC50 > 100,000 nM were labeled as inactive (0).
- Screening Library: A pre-cleaned subset of the Topscience core database (827,897 compounds) was used for high-throughput screening. This library was pre-cleaned according to Lipinski's Rule of Five.
Key Insights and Ideas
- Machine Learning for Efficiency: The XGBoost model achieved a high AUC of 0.9740, demonstrating its effectiveness in rapidly filtering large compound libraries, significantly reducing the screening time (e.g., the ML-based screening of 827,897 compounds took ~20 minutes).
- Hybrid Screening Synergy: Combining LBVS (QSAR) with SBVS (docking) proved more effective than either method alone, improving accuracy and reliability in identifying potential hits.
- Repurposing Existing Drugs: The identification of Rotigotine, a known drug for Parkinson's disease, as a potent BChE inhibitor highlights the potential of repurposing existing medications for AD treatment.
- Dual Mechanism of Inhibition: Piboserod acts as a mixed-type inhibitor, while Rotigotine exhibits non-competitive inhibition, offering different strategies for modulating BChE activity.
- Rotigotine's Stability: Molecular dynamics simulations suggest that Rotigotine forms a more stable complex with BChE compared to piboserod, potentially contributing to its efficacy.
- Importance of ADME/BBB: Integrating in silico ADME and BBB predictions early in the screening process is crucial for identifying candidates with better pharmacokinetic profiles and potential for CNS penetration.
Key Results
- XGBoost Model Performance: Achieved an AUC score of 0.9740, with high precision (96.33%), recall (92.04%), and specificity (97.13%).
- Hit Identification: The hybrid screening strategy identified 12 hits, with piboserod and Rotigotine showing the most potent BChE inhibitory activity.
- Potency:
- Piboserod: IC50 = 15.33 µM
- Rotigotine: IC50 = 12.76 µM
- Safety and Neuroprotection: Piboserod and Rotigotine demonstrated no significant cytotoxicity in PC-12 cells and provided neuroprotective effects against hydrogen peroxide-induced damage.
- Binding Mode: Molecular docking and simulations revealed that both compounds interact with key residues in the BChE active site, including Trp82 and Trp231. Rotigotine's binding appears to be more stable due to interactions with His438 and Trp82.
Limitations and Future Work
- In Vitro vs. In Vivo: While the study shows promising in vitro results, further in vivo studies are needed to confirm the efficacy and safety of these compounds in animal models and eventually in human clinical trials.
- Drug Formulation: The paper mentions the importance of drug formulation aspects, including BBB permeability, but further optimization in this area might be necessary for optimal CNS delivery.
- Mechanism Complexity: While the binding modes are elucidated, the precise downstream effects of BChE inhibition by these specific compounds in the context of AD pathogenesis require further investigation.
- Broader Applicability: The authors suggest that this integrated screening strategy could be valuable for drug discovery targeting other diseases. Future work could involve applying this framework to other therapeutic targets.
Summary updated