검색 상세

Leveraging Machine Learning and Molecular Dynamics for Drug Discovery in COVID-19 and Autoimmune Diseases

초록/요약

Research précis This research advances drug discovery by leveraging cutting-edge computational biology techniques, including machine learning, molecular dynamics, and QSAR modeling, to address therapeutic challenges in inflammatory diseases and viral infections, notably COVID-19. A central focus of the study is the identification of small-molecule inhibitors targeting the IL-1R1/IL-1β interaction, a key regulator in inflammatory pathways. Utilizing transfer learning-enhanced QSAR modeling, predictive frameworks were constructed to identify lead compounds capable of mitigating IL-1β-driven inflammatory responses, presenting potential treatments for conditions such as rheumatoid arthritis. Importantly, the computationally identified inhibitors were experimentally validated, confirming their ability to disrupt the IL-1R1/IL-1β interaction and offering tangible therapeutic prospects. Additionally, the research investigates glycogen synthase kinase-3 (GSK3), a critical protein involved in viral replication and immune regulation, as a therapeutic target against SARS-CoV-2. By integrating molecular dynamics simulations with QSAR-based machine learning, FDA-approved GSK3 inhibitors were repurposed as promising candidates for COVID-19 treatment. The reliability and performance of the computational models were rigorously assessed using benchmarking techniques, demonstrating their predictive accuracy and potential utility in identifying viable therapeutic candidates. Complementary efforts included systematic drug repurposing via virtual screening of FDA-approved compounds against viral targets, coupled with molecular docking and molecular dynamics simulations, which identified additional candidates with potential efficacy against SARS-CoV-2. These findings underscore the efficacy of computational drug discovery methodologies in responding to urgent global health challenges. Collectively, this work highlights the transformative potential of computational tools in streamlining drug discovery and development. By accelerating the identification and validation of novel therapeutics—whether through experimental validation or rigorous computational benchmarks—these approaches provide critical insights into the management of inflammatory diseases and antiviral therapies, with significant implications for addressing pandemics like COVID-19.

more

목차

1.0 Chapter 1: Introduction 2
1.1 Transfer learning QSAR for IL-1R1/IL-1β 2
1.2 GSK3 and RdRp role in SARS-CoV-2 infection 5
1.3 Integrated approaches for inhibitor identification 8
2.0 Chapter 2: Methods 13
2.1 Dataset augmentation and preparation 13
2.1.1 Molecular descriptors 15
2.1.2 Data filtering, splitting, and test selection 15
2.1.3 Transfer and multi-task learning in QSAR modelling 17
2.1.4 QSAR ensemble evaluation methods 18
2.1.5 Virtual screening 19
2.1.6 Molecular docking and molecular dynamics (MD) simulation 19
2.1.7 Post-MD simulation data analysis and visualization 20
2.2 In-vitro validation 21
2.2.1 Culture and differentiation of THP-1 cells 21
2.2.2 Culture of Saos-2 cells 21
2.2.3 Cell viability assay 21
2.2.4 Analysis of cytokine 22
2.2.5 Competitive binding ELISA for IL1R1 inhibitor 22
2.2.6 Osteoclast generation 23
2.2.7 Nitric oxide (NO) assay 24
2.2.8 Western blot analysis 24
2.3 Data compilation and curation 25
2.3.1 Molecular descriptors calculations for GSK3 kinase 26
2.3.2 Data filtering 27
2.3.3 Data splitting and test selection 28
2.3.4 ML-based QSAR classification 29
2.3.5 Statistical assessment for model validation 30
2.3.6 Docking and MD simulation 31
2.3.7 Binding free energy calculations 32
2.3.8 Protein modelling and active-site identification 32
2.4 Library preparation 33
2.4.1 Virtual screening and molecular docking 33
2.5 The MD simulation protocol 34
2.5.1 Ligand topology generation 34
2.5.2 MD simulations of SARS-CoV-2 RdRp 34
2.5.3 Post-MD simulation analysis and visualization 35
2.5.4 Principal component analysis (PCA) 36
2.5.5 The free energy landscape (FEL) 37
2.5.6 The dynamic cross-correlation matrix (DCCM) 37
2.5.7 MM-PBSA calculation and in vitro validation 38
3.0 Chapter 3: Results and discussion 42
3.1 Insights on IL-1R1/IL-1β inhibitors 42
3.1.1 Exploring the diversity of IL1-R1 inhibitors: chemical space analysis 42
3.1.2 Transfer learning and ensemble model construction 47
3.1.3 Virtual screening with ensemble QSAR modelling 50
3.2 In-vitro screening 53
3.2.1 Lethal dose (LD50) and inhibitory concentration (IC50) calculations 54
3.2.2 Anti-IL1-R1 activity osteoclast: ELISA and NO assay 57
3.2.3 IRI-1 inhibition of IL1-R1: western blot and competitive binding ELISA 59
3.3 MD simulation of IL1-R1-ligand complexes 60
3.3.1 RMSD reveals ligand-induced conformational changes 62
3.3.2 RMSF analysis of ligand-modulated protein dynamics 65
3.3.3 Radius of gyration (Rg) analysis of ligand-induced compaction 66
3.3.4 Hydrogen bond analysis of ligand-protein interactions 68
3.3.5 Unveiling protein-ligand binding dynamics 68
3.3.6 Non-bonded interaction analysis 72
3.4 Conclusion 72
3.5 Machine learning for GSK3 inhibitors 76
3.5.1 Exploratory chemical space analysis of GSK3 inhibitors 77
3.5.2 Ensemble boosting model GSK3 kinase inhibitors 81
3.5.3 Atomic-level characterization and binding free energy calculations 86
3.6 Conclusion 91
3.7 Virtual screening and molecular docking 94
3.7.1 Analysis of the interaction of top hits with RdRp 95
3.7.2 Cα atoms variations in RdRp with or without drug binding 97
3.7.3 Calculation of equilibrium conformation of systems 99
3.7.4 Analysis of essential dynamics of protein 100
3.7.5 Exploration of protein folding dynamics 102
3.7.6 Time-correlated protein domain motions and flexibility 103
3.7.7 Binding free energy calculations 104
3.7.8 In vitro antiviral activity 106
3.8 Conclusion 108
References 109

more
반출 Meta View 목록