Multiscale computational elucidation of structure-function relationships in chemical and biomolecular systems

다국어 초록 (Multilingual Abstract)

Understanding how molecular structure determines function is a central challenge in chemistry, materials science, and molecular biology. Essentially, functional behavior originates from variations in atomic and electronic structure that reshape the potential energy landscape to govern molecular function. While structure–function relationships are universal, meaningful characterization requires examining structural causality at a computational resolution appropriate to the size and nature of the system under investigation. In this dissertation, a multiscale computational perspective is employed to analyze these relationships across diverse chemical and biomolecular systems.
At the level of organic reaction chemistry, density functional theory calculations are used to explore potential energy surfaces through geometry optimization, transition state identification, and reaction path analysis, thereby delineating core reaction mechanisms. When full mechanistic enumeration is impractical due to complexity or competing pathways, the analytical focus shifts to comparative analysis of optimized stable structures, enabling identification of key electronic and geometric motifs governing product formation.
For amorphous organic semiconductor materials, representative local structures formed through intermolecular interactions are validated by time-dependent density functional theory calculations and experimental spectroscopy data, confirming that interaction-induced polarization provides a structural origin for enhanced physical and electrical properties.
For functions such as sweet taste perception that emerge at macroscopic or physiological levels across large populations of diverse molecular structures, machine-learning-based feature analysis is used to identify structurally significant features associated with function. For biomolecular systems with experimentally elusive structures, molecular dynamics simulations provide critical insights, establishing that function originates from ligand-dependent conformational ensembles instead of a single static state.
Overall, this dissertation demonstrates that a single universal principle of structure–function relationships can be systematically elucidated across chemical and biomolecular systems by analyzing structural causality at computational resolution tailored to the dominant form of structural variation.

국문 초록 (Abstract)

분자 구조가 어떻게 특정한 기능을 만들어내는지 이해하는 일은 화학과 분자생물학, 재료과학을 관통하는 핵심 과제이다. 분자의 모든 기능은 결국 원자와 전자의 미세한 배치 변화에서 시작되며, 이것이 퍼텐셜 에너지 지형을 뒤바꾸어 최종적인 거동을 결정하기 때문이다. 이러한 구조–기능 관계의 원리는 어디에나 적용되지만, 이를 정확히 파악하기 위해서는 연구 대상의 규모와 성격에 맞추어 계산 해상도를 최적화하는 과정이 반드시 필요하다. 본 논문에서는 다중 스케일 계산 접근법을 도입하여, 다양한 시스템 속에 숨겨진 구조적 인과관계를 체계적으로 분석하고자 하였다.
먼저 유기 반응 연구에서는 밀도범함수 이론을 활용해 구조 최적화와 전이 상태를 추적함으로써 반응 메커니즘의 핵심을 규명했다. 특히 반응 경로가 너무 복잡하여 모든 과정을 일일이 나열하기 어려운 경우에는, 에너지 장벽을 계산하는 기존의 방식에서 벗어나 반응에 최적으로 배열된 활성 중간체들의 구조를 분석하는 방식으로 분석의 틀을 전환하였다. 이를 통해 메커니즘 추적 대신 로컬 최솟값 구조들의 전하와 전자밀도를 비교하여 생성물 형성을 결정짓는 핵심적인 전자적·기하학적 요소를 파악했다.
비정질 유기 반도체 분야에서는 분자 간 상호작용으로 형성된 구조를 시간 의존 밀도범함수 이론 계산과 실측된 실험 분광 데이터로 검증하였다. 이 과정에서 리튬 결합에 의한 분극 현상이 재료의 전기적·물리적 성능을 높이는 근본 원인임을 확인했다. 더 나아가, 수백 개가 넘는 분자의 구조에서 공통 기능을 확인해야 하는 단맛 지각과 같은 생리학적 현상은 기계학습 기반의 특성 분석을 통해 기능과 연결된 유의미한 구조적 특징을 추출해냈다. 또한, 실험적으로 구조를 알기 어려운 생체 시스템의 경우 분자동역학 시뮬레이션을 수행하여, 단일 구조가 아닌 리간드에 따라 변화하는 구조 앙상블이 기능을 결정한다는 사실을 밝혔다.
결론적으로 본 학위논문은 구조–기능 관계라는 보편적인 원리가 각 문제에 적합한 계산 해상도로 분석될 때, 복잡한 화학 및 생체 시스템 전반에서 얼마나 체계적으로 설명될 수 있는지를 보여준다.

목차 (Table of Contents)

• Chapter 1. Introduction 1
• 1.1. Structure–Function Relationships in Molecular Systems 1
• 1.2. Structural Causality and the Role of Computational Resolution 2
• 1.3. Multiscale Computational Approaches to Structure–Function Problems 5
• 1.4. Scope and Organization of This Dissertation 6
• Chapter 2. Density Functional Theory Study of LiOtBu-Promoted Deconjugation of α,β-Unsaturated Diesters 10
• 2.1. Introduction 10
• 2.2. Computational Details 13
• 2.2.1. Electronic Structure Methods and Basis Sets 13
• 2.2.2. Geometry Optimization and Solvation Model 14
• 2.2.3. Transition State Search and Reaction Pathway Verification 15
• 2.2.4. Thermodynamic Contributions 16
• 2.3. Results and Discussion 16
• 2.3.1. Lithium Coordination and Ground-State Organization 18
• 2.3.2. γ-Deprotonation Pathways from the Lithium-Coordinated Reactant 21
• 2.3.3. Quenching and α-Protonation of the Dienolate Intermediate 24
• 2.3.4. Comparison with Non-Coordinating Base Conditions 27
• 2.4. Chapter Summary 30
• Chapter 3. Structure-Centered Computational Study of Superacid-Catalyzed Reductive Amination Enabled by Pinacolborane Assistance 32
• 3.1. Introduction 32
• 3.2. Computational Details 36
• 3.2.1. Geometry Optimization and Frequency Analysis 37
• 3.2.2. Molecular Electrostatic Potential Calculations 38
• 3.2.3. Single-Point Calculations and Electronic Structure Analysis 38
• 3.3. Results and Discussion 39
• 3.3.1. Initial Activation of Pinacolborane under Superacidic Conditions 42
• 3.3.2. Carbonyl Activation in the Presence of the HBpin–Tf2NH 44
• 3.3.3. Response of Carbamate within HBpin–Tf2NH 48
• 3.3.4. Implications for the Reaction Initiation 51
• 3.4. Chapter Summary 52
• Chapter 4. Local Structure–Function Relationships in Amorphous Organic Semiconductors Revealed by Intermolecular Lithium Bonding 54
• 4.1. Introduction 54
• 4.2. Computational Details 58
• 4.2.1. Ground State Electronic Structure Analysis 59
• 4.2.2. Excited State Calculations 59
• 4.2.3. Molecular Volume Calculations 60
• 4.3. Results and Discussion 60
• 4.3.1. UV–Vis Spectroscopic Signatures of Lithium Participation in Intermolecular Interactions 63
• 4.3.2. Infrared Identification of Oxygen-Centered Lithium Bonding 67
• 4.3.3. Electronic Polarization and Dipole Formation Induced by Lithium–Oxygen Bonding 70
• 4.3.4. Volume Ratio Dependent Formation of Lithium-Bonded Complexes and the Origin of the Optimal 50 vol% Composition 74
• 4.4. Chapter Summary 78
• Chapter 5. Machine-Learning-Based Elucidation of Structure–Function Relationships in Sweet Taste Perception 81
• 5.1. Introduction 81
• 5.2. Computational Details 85
• 5.2.1. Dataset Construction and Curation 85
• 5.2.2. Machine Learning Models and Implementation 87
• 5.2.3. Model Evaluation Metrics 89
• 5.3. Results and Discussion 90
• 5.3.1. Sweetness Prediction Models and Model Performance 90
• 5.3.2. Quantitative Comparison with Existing Models and Robustness of Prediction 93
• 5.3.3. Structural Features Governing Sweet Taste Perception 95
• 5.3.4. Decision Logic and Threshold-Based Rules 97
• 5.3.5. Integrated Structure–Function Interpretation 101
• 5.4. Chapter Summary 103
• Chapter 6. Ensemble-Based Structural Elucidation of Ligand-Dependent Modulation in the Ectopic Olfactory Receptor OR2AT4 106
• 6.1. Introduction 106
• 6.2. Computational Details 111
• 6.2.1. Sequence Conservation Analysis and Binding-Pocket Definition 111
• 6.2.2. Homology Modeling and System Preparation 112
• 6.2.3. Molecular Dynamics Simulation Protocol 113
• 6.2.4. Apo-State Clustering Analysis 114
• 6.2.5. Ligand Preparation and Docking 116
• 6.2.6. Trajectory Analysis 116
• 6.3. Results and Discussion 117
• 6.3.1. Structural Validity of the OR2AT4 model and Conserved GPCR features 117
• 6.3.2. Apo-State Dynamics Define a Stable Structural Baseline for Ligand-Induced Effects 119
• 6.3.3. Docking Distinguishes Agonist, Antagonist, and Inactive Ligands by Binding Compatibility 122
• 6.3.4. Holo-State Molecular Dynamics Simulations Reveal Ligand-Dependent Binding Stability 125
• 6.3.5. Ligand-Induced Reorganization of Local Interaction Networks Centered on GLU116 129
• 6.4. Integrated interpretation of ligand-dependent modulation in OR2AT4 132
• 6.4.1. Ensemble-Based Perspective on OR2AT4 Modulation 132
• 6.4.2. Redistribution of Local Interaction Networks as a Regulatory Mode 133
• 6.4.3. TM6-associated Coupling and Cytoplasmic Structure 134
• 6.4.4. Scope and Interpretative Limits 136
• 6.5. Chapter Summary 137
• Chapter 7. Conclusion 141
• References 146
• 논문요약 175