graph LR
Core_Generation_Runner["Core Generation Runner"]
Chemical_Structure_Utilities["Chemical Structure Utilities"]
Conformation_Evaluation["Conformation Evaluation"]
Distance_Geometry_Operations["Distance Geometry Operations"]
PyTorch_Helper_Utilities["PyTorch Helper Utilities"]
Molecular_Data_Preparation["Molecular Data Preparation"]
Core_Generation_Runner -- "utilizes" --> Molecular_Data_Preparation
Core_Generation_Runner -- "leverages" --> PyTorch_Helper_Utilities
Core_Generation_Runner -- "interacts with" --> Distance_Geometry_Operations
Core_Generation_Runner -- "interacts with" --> Conformation_Evaluation
Core_Generation_Runner -- "interacts with" --> Chemical_Structure_Utilities
Chemical_Structure_Utilities -- "depends on" --> PyTorch_Helper_Utilities
Conformation_Evaluation -- "depends on" --> Chemical_Structure_Utilities
Conformation_Evaluation -- "depends on" --> PyTorch_Helper_Utilities
Distance_Geometry_Operations -- "depends on" --> PyTorch_Helper_Utilities
Molecular_Data_Preparation -- "utilizes" --> Distance_Geometry_Operations
Molecular_Data_Preparation -- "utilizes" --> Chemical_Structure_Utilities
The Conformation Generation & Evaluation subsystem is responsible for creating 3D molecular conformations from distance matrices using distance geometry and assessing their quality. It integrates core generation runners that orchestrate Langevin Dynamics simulations for both position and distance-based approaches, leveraging PyTorch utilities for numerical stability and data handling. The generated conformations are then evaluated using metrics like RMSD and Maximum Mean Discrepancy (MMD), with essential chemical structure utilities supporting molecular data manipulation and analysis throughout the process. Molecular data preparation, including SMILES to data conversion, forms the initial step in this pipeline.
This component orchestrates the molecular conformation generation process, including Langevin Dynamics simulations for both position and distance-based approaches. It serves as the primary interface for generating samples from SMILES strings or test sets.
Related Classes/Methods:
-
ConfGF.confgf.runner.default_runner.DefaultRunner.position_Langevin_Dynamics(222:258) -
ConfGF.confgf.runner.default_runner.DefaultRunner.ConfGF_generator(260:280) -
ConfGF.confgf.runner.default_runner.DefaultRunner.ConfGFDist_generator(283:304) -
ConfGF.confgf.runner.default_runner.DefaultRunner.generate_samples_from_smiles(307:351) -
ConfGF.confgf.runner.default_runner.DefaultRunner.generate_samples_from_testset(354:395) -
ConfGF.confgf.runner.default_runner.DefaultRunner.convert_score_d(187:191) -
ConfGF.confgf.runner.default_runner.DefaultRunner.distance_Langevin_Dynamics(195:218)
Provides essential functions for manipulating and analyzing chemical structures, such as setting atom positions in RDKit molecules, calculating RMSD, and retrieving atom symbols. These utilities are fundamental for handling molecular data.
Related Classes/Methods:
This component is responsible for evaluating the quality and diversity of generated molecular conformations. It includes functions for computing RMSD confusion matrices and evaluating distance-based metrics like Maximum Mean Discrepancy (MMD) using Gaussian kernels.
Related Classes/Methods:
Handles operations related to molecular distance geometry, including embedding 3D coordinates from distance matrices and calculating distance matrices from atomic positions. These are crucial for working with molecular conformations in a distance space.
Related Classes/Methods:
Contains general utility functions that leverage PyTorch, such as clipping tensor norms and repeating data, which are commonly used in deep learning contexts to ensure numerical stability and data preparation.
Related Classes/Methods:
Manages the conversion of chemical identifiers, specifically SMILES strings, into structured data formats suitable for processing by the molecular generation models.
Related Classes/Methods: