Skip to content

Latest commit

 

History

History
220 lines (190 loc) · 15.3 KB

File metadata and controls

220 lines (190 loc) · 15.3 KB
graph LR
    Data_Pipeline["Data Pipeline"]
    Model_Core_Training_Frameworks["Model Core & Training Frameworks"]
    Specialized_Model_Architectures["Specialized Model Architectures"]
    Molecular_Chemoinformatics_Utilities["Molecular & Chemoinformatics Utilities"]
    Experimentation_Evaluation["Experimentation & Evaluation"]
    Data_Pipeline -- "provides processed data to" --> Model_Core_Training_Frameworks
    Data_Pipeline -- "prepares input for" --> Model_Core_Training_Frameworks
    Data_Pipeline -- "provides processed data to" --> Specialized_Model_Architectures
    Data_Pipeline -- "prepares input for" --> Specialized_Model_Architectures
    Data_Pipeline -- "provides datasets for" --> Experimentation_Evaluation
    Data_Pipeline -- "supplies data to" --> Experimentation_Evaluation
    Model_Core_Training_Frameworks -- "provides base for" --> Specialized_Model_Architectures
    Model_Core_Training_Frameworks -- "implements core logic for" --> Specialized_Model_Architectures
    Model_Core_Training_Frameworks -- "is evaluated by" --> Experimentation_Evaluation
    Model_Core_Training_Frameworks -- "provides models for" --> Experimentation_Evaluation
    Specialized_Model_Architectures -- "is evaluated by" --> Experimentation_Evaluation
    Specialized_Model_Architectures -- "provides predictions to" --> Experimentation_Evaluation
    Specialized_Model_Architectures -- "leverages for chemical operations" --> Molecular_Chemoinformatics_Utilities
    Specialized_Model_Architectures -- "uses for molecular data handling" --> Molecular_Chemoinformatics_Utilities
    Molecular_Chemoinformatics_Utilities -- "supports featurization in" --> Data_Pipeline
    Molecular_Chemoinformatics_Utilities -- "assists in data preparation for" --> Data_Pipeline
    Molecular_Chemoinformatics_Utilities -- "provides chemical insights to" --> Specialized_Model_Architectures
    Molecular_Chemoinformatics_Utilities -- "supports molecular computations for" --> Specialized_Model_Architectures
    Experimentation_Evaluation -- "accesses data from" --> Data_Pipeline
    Experimentation_Evaluation -- "relies on data from" --> Data_Pipeline
    Experimentation_Evaluation -- "optimizes" --> Model_Core_Training_Frameworks
    Experimentation_Evaluation -- "benchmarks" --> Model_Core_Training_Frameworks
    Experimentation_Evaluation -- "evaluates" --> Specialized_Model_Architectures
    Experimentation_Evaluation -- "tunes" --> Specialized_Model_Architectures
    click Data_Pipeline href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//deepchem/Data Pipeline.md" "Details"
    click Model_Core_Training_Frameworks href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//deepchem/Model Core & Training Frameworks.md" "Details"
    click Specialized_Model_Architectures href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//deepchem/Specialized Model Architectures.md" "Details"
    click Molecular_Chemoinformatics_Utilities href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//deepchem/Molecular & Chemoinformatics Utilities.md" "Details"
    click Experimentation_Evaluation href "https://github.com/CodeBoarding/GeneratedOnBoardings/blob/main//deepchem/Experimentation & Evaluation.md" "Details"
Loading

CodeBoardingDemoContact

Component Details

The DeepChem library provides a comprehensive platform for cheminformatics and machine learning in drug discovery, materials science, and quantum chemistry. Its core functionality revolves around a robust data pipeline that handles data loading, featurization, and preprocessing. This prepared data then feeds into a flexible model core that supports various deep learning frameworks and a wide array of specialized model architectures, including graph neural networks, generative models, and quantum chemistry models. Complementing these are extensive molecular and chemoinformatics utilities for detailed chemical manipulations and interactions, and a dedicated experimentation and evaluation component that facilitates benchmarking, hyperparameter tuning, and meta-learning to optimize model performance and accelerate scientific discovery.

Data Pipeline

Manages the entire data lifecycle, from loading raw data and converting it into numerical features (featurization) to applying various transformations (preprocessing) and splitting datasets for training and evaluation. It handles diverse data formats and ensures data readiness for machine learning models.

Related Classes/Methods:

  • deepchem.data.datasets.Dataset (221:743)
  • deepchem.data.data_loader.DataLoader (72:278)
  • deepchem.utils.data_utils (full file reference)
  • deepchem.utils.cache_utils (full file reference)
  • deepchem.feat.base_classes (full file reference)
  • deepchem.feat.deepvariant_featurizer (full file reference)
  • deepchem.feat.graph_data (full file reference)
  • deepchem.feat.reaction_featurizer (full file reference)
  • deepchem.feat.binding_pocket_features (full file reference)
  • deepchem.feat.dft_data (full file reference)
  • deepchem.feat.mol_graphs (full file reference)
  • deepchem.feat.roberta_tokenizer (full file reference)
  • deepchem.feat.atomic_conformation (full file reference)
  • deepchem.feat.smiles_tokenizer (full file reference)
  • deepchem.feat.graph_features (full file reference)
  • deepchem.feat.deepvariant_pileup_featurizer (full file reference)
  • deepchem.feat.sequence_featurizers (full file reference)
  • deepchem.feat.molecule_featurizers (full file reference)
  • deepchem.feat.material_featurizers (full file reference)
  • deepchem.feat.complex_featurizers (full file reference)
  • deepchem.feat.vocabulary_builders (full file reference)
  • deepchem.trans.transformers.Transformer (56:238)
  • deepchem.trans.duplicate (full file reference)
  • deepchem.splits.splitters.Splitter (31:325)
  • deepchem.splits.task_splitter (full file reference)
  • deepchem.data.supports (full file reference)
  • deepchem.contrib.atomicconv.splits (full file reference)

Model Core & Training Frameworks

Provides the foundational infrastructure for building, training, and managing machine learning models. This includes base model classes, optimizers, loss functions, and integrations with deep learning frameworks like Keras/TensorFlow, PyTorch, and JAX, enabling flexible model development and deployment.

Related Classes/Methods:

Specialized Model Architectures

Houses a diverse collection of pre-built machine learning models tailored for specific scientific domains and data types, including Graph Neural Networks (GNNs), sequence models, image models, generative models, reinforcement learning agents, quantum chemistry models (DFT), and traditional/ensemble methods. These models leverage the core training infrastructure to solve complex problems in chemistry and biology.

Related Classes/Methods:

  • deepchem.models.graph_models (full file reference)
  • deepchem.models.torch_models.graphconvmodel (full file reference)
  • deepchem.models.torch_models.mpnn (full file reference)
  • deepchem.models.torch_models.gat (full file reference)
  • deepchem.models.torch_models.gcn (full file reference)
  • deepchem.models.torch_models.attentivefp (full file reference)
  • deepchem.models.torch_models.pagtn (full file reference)
  • deepchem.models.torch_models.infograph (full file reference)
  • deepchem.models.torch_models.megnet (full file reference)
  • deepchem.models.torch_models.cgcnn (full file reference)
  • deepchem.models.torch_models.lcnn (full file reference)
  • deepchem.models.torch_models.gnn3d (full file reference)
  • deepchem.models.torch_models.mxmnet (full file reference)
  • deepchem.models.torch_models.dmpnn (full file reference)
  • deepchem.models.torch_models.acnn (full file reference)
  • deepchem.models.torch_models.grover (full file reference)
  • deepchem.models.chemnet_models (full file reference)
  • deepchem.models.seqtoseq (full file reference)
  • deepchem.models.text_cnn (full file reference)
  • deepchem.models.torch_models.smiles2vec (full file reference)
  • deepchem.models.torch_models.text_cnn (full file reference)
  • deepchem.models.torch_models.hf_models (full file reference)
  • deepchem.models.torch_models.antibody_modeling (full file reference)
  • deepchem.models.torch_models.chemberta (full file reference)
  • deepchem.models.torch_models.molformer (full file reference)
  • deepchem.models.torch_models.oneformer (full file reference)
  • deepchem.models.torch_models.inceptionv3 (full file reference)
  • deepchem.models.torch_models.mobilenetv2 (full file reference)
  • deepchem.models.torch_models.unet (full file reference)
  • deepchem.models.torch_models.cnn (full file reference)
  • deepchem.models.molgan (full file reference)
  • deepchem.models.normalizing_flows (full file reference)
  • deepchem.models.torch_models.flows (full file reference)
  • deepchem.models.torch_models.gan (full file reference)
  • deepchem.models.torch_models.molgan (full file reference)
  • deepchem.models.torch_models.lstm_generator_models (full file reference)
  • deepchem.rl (full file reference)
  • deepchem.contrib.rl (full file reference)
  • deepchem.models.dft (full file reference)
  • deepchem.models.torch_models.ferminet (full file reference)
  • deepchem.models.sklearn_models (full file reference)
  • deepchem.models.gbdt_models (full file reference)
  • deepchem.models.fcnet (full file reference)
  • deepchem.models.robust_multitask (full file reference)
  • deepchem.models.scscore (full file reference)
  • deepchem.models.IRV (full file reference)
  • deepchem.contrib.autoencoder_models (full file reference)
  • deepchem.contrib.dragonn (full file reference)
  • deepchem.contrib.DeepMHC (full file reference)
  • deepchem.contrib.laplacian (full file reference)
  • deepchem.contrib.DiabeticRetinopathy (full file reference)
  • deepchem.contrib.vina_model (full file reference)
  • deepchem.contrib.atomicconv.models (full file reference)
  • deepchem.contrib.hagcn (full file reference)
  • deepchem.contrib.one_shot_models (full file reference)
  • deepchem.contrib.mpnn (full file reference)

Molecular & Chemoinformatics Utilities

Provides a comprehensive set of fundamental helper functions and tools for handling molecular structures, chemical properties, geometric computations, and general scientific data manipulation. This includes RDKit operations, coordinate manipulation, non-covalent interaction analysis, conformer generation, PDBQT handling, and specialized utilities for polymers, genomics, DFT, and molecular docking.

Related Classes/Methods:

  • deepchem.utils.rdkit_utils (full file reference)
  • deepchem.utils.geometry_utils (full file reference)
  • deepchem.utils.coordinate_box_utils (full file reference)
  • deepchem.utils.fragment_utils (full file reference)
  • deepchem.utils.noncovalent_utils (full file reference)
  • deepchem.utils.conformers (full file reference)
  • deepchem.utils.pdbqt_utils (full file reference)
  • deepchem.utils.molecule_feature_utils (full file reference)
  • deepchem.utils.poly_wd_graph_utils (full file reference)
  • deepchem.utils.poly_converters (full file reference)
  • deepchem.utils.grover (full file reference)
  • deepchem.utils.electron_sampler (full file reference)
  • deepchem.utils.sequence_utils (full file reference)
  • deepchem.utils.voxel_utils (full file reference)
  • deepchem.utils.genomics_utils (full file reference)
  • deepchem.utils.dft_utils (full file reference)
  • deepchem.utils.docking_utils (full file reference)
  • deepchem.dock (full file reference)
  • deepchem.feat.complex_featurizers (full file reference)

Experimentation & Evaluation

Facilitates the entire experimental workflow, from accessing standardized datasets (MolNet) and benchmarking models to evaluating performance using various metrics and optimizing model hyperparameters. It also includes advanced learning techniques like meta-learning to enhance model adaptability and efficiency.

Related Classes/Methods:

  • deepchem.molnet (full file reference)
  • deepchem.datasets.construct_pdbbind_df (full file reference)
  • deepchem.molnet.load_function (full file reference)
  • deepchem.molnet.run_benchmark (17:237)
  • deepchem.contrib.pubchem_dataset (full file reference)
  • deepchem.metrics.Metric (421:726)
  • deepchem.metrics.genomic_metrics (full file reference)
  • deepchem.metrics.score_function (full file reference)
  • deepchem.utils.evaluate (full file reference)
  • deepchem.hyper (full file reference)
  • deepchem.hyper.base_classes (full file reference)
  • deepchem.hyper.gaussian_process (full file reference)
  • deepchem.hyper.grid_search (full file reference)
  • deepchem.hyper.random_search (full file reference)
  • deepchem.metalearning (full file reference)
  • deepchem.metalearning.maml (full file reference)
  • deepchem.metalearning.torch_maml (full file reference)