검색 상세

Dynamic Insights into Protein- Protein Interaction Modulation By Small Molecules

초록/요약

Targeting macromolecular protein-protein interactions (PPIs) with small-molecule inhibitors remains a formidable challenge in drug discovery due to the expansive and flat nature of the binding interfaces. This thesis presents the development and application of two distinct, AI-augmented discovery pipelines to identify novel inhibitors for pivotal signaling axes in oncology and immunology: the PD-1/PD-L1 immune checkpoint and the IL23/IL23R inflammatory pathway. In the first study, a machine learning-based quantitative structure-activity relationship (ML-QSAR) model was integrated with high-throughput virtual screening and molecular docking to target the PD-L1 dimer interface. Extensive molecular dynamics (MD) simulations elucidated the atomic-level disruption mechanism of the lead candidate, PDA13, which was validated in vitro to inhibit the PD-1/PD-L1 interaction with an IC50 of 17.53 µM. In the second study, a deep generative model-Sequential Attachment-based Fragment Embedding (SAFE)-was employed to design novel scaffolds targeting the IL23(p19) subunit. Following virtual screening and cellular validation using HEK-Blue reporter cells, the lead compound Inh-31 was identified. Advanced all-atom MD simulations (300 ns) revealed that Inh-31 functions via a "conformational locking" mechanism, inducing allosteric stabilization and global rigidification of the IL23R subunit (RMSD < 0.5 nm), thereby preventing the assembly of a signaling-competent complex and inhibiting the downstream JAK-STAT3 cascade. Collectively, these projects demonstrate the synergistic power of coupling deep learning, generative chemistry, and biophysical simulations with experimental validation. The results provide a robust framework for overcoming the inherent difficulties of PPI modulation, offering promising scaffolds for the next generation of immunotherapeutic and anti-inflammatory agents.

more

목차

Project 1 1
Summary 2
1. Introduction 4
2. Methodology 7
2.1 Virtual Screening Pipeline and Hit Identification 7
2.2 Data Compilation and Classification 7
2.3 Molecular Descriptor Generation and Analysis 8
2.3.1 Descriptor Calculation and Physicochemical Space 8
2.3.2 Univariate and Multivariate Analysis 8
2.4 Structure-Activity Relationship (SAR) Exploration 8
2.4.1 Structure-Activity Similarity (SAS) Maps 8
2.4.2 Structure-Activity Landscape Index (SALI) 9
2.4.3 Molecular Descriptor Generation and Feature Selection for QSAR 9
2.4.4 Feature Selection 9
2.4.5 QSAR Model Construction and Validation 9
2.4.6 QSAR Model Construction: Binary Classification 10
2.4.7 Internal Validation 10
2.4.8 External Validation: Testing on New Data 10
2.4.9 Ensuring Model's Applicability 11
2.5 Chemotype Analysis 11
2.5.1 Scaffold and R-Group Analysis 11
2.6 Software and Data Visualization 12
2.7 Preparation of Virtual Screening Library 12
2.8 Fingerprint-Based Structure Similarity 12
2.9 Molecular Docking: Protein Structure and Ligand Preparation 12
2.10 ADMET Analysis of Potential Inhibitors 12
2.11 Molecular Dynamics (MD) Simulation and Post-MD Analyses 13
2.12 In Vitro Analysis 13
2.12.1 Compound Screening and IC50 Determination 13
2.12.2 Cell Lines and Culture 13
2.12.3 Cell Viability Assay 14
2.12.4 PD-1/PD-L1 Blockade Assay 14
2.12.5 Surface Plasmon Resonance (SPR) assay 14
3. Results 17
3.1 Data Exploration: PD-1/PD-L1 Inhibitor Characteristics 17
3.2 Visualizing PD-1/PD-L1 Inhibitor Diversity: PCA Analysis 18
3.3 Structure-Activity Landscape Index (SALI) Analysis 19
3.4 Building Predictive Models for PD-1/PD-L1 Inhibitor Activity: Ranking the Algorithms and Hyperparameter Tuning 20
3.5 External Model Validation and Assessment 21
3.6 PD-1/PD-L1 Inhibitor Scaffolds: Diversity, Analysis, and Optimization 22
3.7 R-group Analysis 24
3.8 Structural Dynamics of the PD-L1 Dimer Interface 28
3.9 RMSD Analysis – Stability and Structural Deviations 28
3.10 Radius of Gyration (Rg) Analysis – Equilibration and Compactness 28
3.11 Analysis of Interface Dynamics: RMSF and Interchain Synchronization 29
3.12 Ligand-Induced Conformational Dynamics and Rationale for PDA13 Selection 29
3.13 MM-PBSA and Decomposition Analysis 31
3.14 In Vitro PD-1/PD-L1 Blockade Assay 32
4. Discussion 37
5. Conclusion 42
Project 2 43
Summary 44
1. Introduction 46
2. Methodology 49
2.1 Deep Generative and Constrained Molecular Design 49
2.2 Scaffold-Conditioned Generation Using the SAFE Algorithm 49
2.3 DeepPurpose for Drug-Target Interaction Prediction 50
2.4 Virtual Screening Library Preparation 51
2.5 Fingerprint-Based Similarity Screening 52
2.6 Protein Structure Preparation and Docking 52
2.7 ADMET Analysis 52
2.8 Molecular Dynamics Simulation 53
2.9 Cell Culture and In Vitro Assays 53
2.9.1 SEAP Assay 53
2.9.2 Cytotoxicity of the Compounds 53
3. Results 55
3.1 DTI Prediction and Lead Identification 55
3.2 Structural Insights into Binding Mode 56
3.3 MD Simulation of Inhibitor-Induced Conformational Changes 58
3.3.1 Global Stability and Structural Compactness 58
3.3.2 Subunit-Specific Dynamics and Allosteric Stabilization 58
3.3.3 Residue-Specific Flexibility (RMSF) 58
3.4 Thermodynamic Analysis of Inhibitor Binding (MM/PBSA) 59
3.5 Structural Insights from Superimposition 61
3.6 Comparative Analysis of Interfacial Contacts: "Apo" vs. Inhibitor-Bound Complex 62
3.6.1 The Native IL23/IL23R Interaction ("Apo" Form) 62
3.6.2 Mechanism of Inhibition by Small Molecule 63
3.6.3 Retention and Non-Native Contacts 63
3.7 Evaluation of Small Molecule Inhibitors on IL23/IL23R Signaling 65
3.7.1 Initial Screen for Efficacy and Toxicity 65
3.7.2 Concentration-Dependent Efficacy and IC50 Determination 65
3.7.3 Cytotoxicity Profile 65
4. Discussion 69
5. Conclusion 73
References 74
Supplementary Data – Project 1 82
Supplementary Data – Project 2 97

more
반출 Meta View 목록