deepchem/deepchem

Features

• Molecular Featurization Libraries - Converts molecular structures into numerical representations for machine learning models.
• Molecular - Provides a complete toolchain for applying deep learning to molecular and chemical data.
• Few-Shot Learning Mechanisms - Implements few-shot learning mechanisms for fast adaptation across molecular tasks with limited data.
• Model Fine-Tuning - Supports fine-tuning pre-trained meta-models with gradient descent steps for task specialization.
• Meta Learning Frameworks - Encapsulates models and task data into a reusable meta-learning interface for few-shot training.
• Task Encapsulation Interfaces - Encapsulates tasks and models for few-shot meta-learning with fast adaptation across many tasks.
• Molecular Docking Pipelines - Automates molecular docking by chaining pocket detection, pose generation, and scoring into a single pipeline.
• Binding Pose Generators - Compute energetically favorable conformations of a small molecule interacting with a protein by invoking external docking software.
• Docking Pose Scorers - Evaluate the quality of predicted binding poses by computing interaction terms such as repulsion, hydrophobic contacts, and hydrogen bonds.
• Molecular Tokenization - Implements subword tokenization of SMILES strings for transformer-based molecular language models.
• Binding Affinity Prediction - Predicts protein-ligand binding affinity and docking poses for structure-based drug design.
• Computational Workbenches - Provides a comprehensive toolkit for molecular modeling, docking, and property prediction in drug discovery.
• Molecular Deep Learning Frameworks - Applies deep learning to molecular, chemical, and biological data for drug discovery and materials science.
• Property Forecasting Models - Applies ML to structural data to forecast physical characteristics of materials.
• Property Prediction - Predicts physical and chemical properties of molecular structures from their data.
• Sequence Featurization - Featurizes genomic and proteomic sequences from alignment files for downstream machine learning models.
• Disk-Backed Stores - Provides disk-backed storage for molecular datasets too large to fit in memory.
• Molecular Workflow Orchestration - Orchestrates deep learning workflows on molecular and chemical data for drug discovery.
• Molecular - Stores molecules as canonicalized graphs with atom features and adjacency lists for graph neural network models.
• Deep Learning Toolkits - Provides tools for featurizing molecules and training neural networks for chemical property prediction.
• Molecular Property Prediction - Trains ML models on molecular data to predict chemical and quantum properties.
• Crystal Property Forecasting - Forecasts physical and chemical properties of materials from crystal structural data using ML.
• Featurization Pipelines - Converts molecular structures into fingerprints, graphs, and descriptors for ML model input.
• Materials Forecasting Platforms - Trains and evaluates ML models to forecast physical and chemical properties of materials.
• Affinity Scoring Functions - Computes binding affinity scores between small molecules and protein targets using learned representations.
• Complex Featurization Pipelines - Provides featurization pipelines that convert protein-ligand complexes into numerical descriptors for ML models.
• Molecular Graph Representations - Ships weighted directed graph representations of macromolecules for polymer and molecular modeling.
• Molecular Featurization Pipelines - Provides a unified featurizer-pipeline abstraction that decouples molecular data loading from model training.
• Machine Learning Workflow Libraries - Ships a complete workflow library for molecular machine learning from data loading through evaluation.
• One-Hot - Converts arbitrary strings or SMILES into fixed-length one-hot encoded arrays for sequence-based models.
• Molecular SMICS Tokenizers - Splits SMILES strings into subword tokens using a learned vocabulary for transformer-based molecular models.
• Cell Counting in Microscopy Images - Provides deep learning-based cell segmentation and counting from microscopy images.
• In-Memory Data Stores - Holds features, labels, weights, and identifiers as NumPy arrays in RAM for fast access and iteration.
• Dataset Iterators - Provides a mechanism for yielding minibatches of features, labels, weights, and identifiers from a Dataset.
• Sharded NumPy Dataset Persisters - Provides sharded NumPy file persistence for datasets too large to fit in RAM.
• Multi-Framework Dataset Converters - Provides a common interface for converting datasets between NumPy, pandas, TensorFlow, and PyTorch formats.
• Multi-Framework Dataset Conversions - Converts datasets between NumPy, pandas, TensorFlow, and PyTorch formats through a common interface.
• CSV Data Loaders - Reads tabular data from CSV files, applies a featurizer, and stores the result as a Dataset.
• Molecular Graph Stores - Stores molecules as canonicalized graphs with atom features and adjacency lists for GNN models.
• FASTA File Loaders - Reads nucleotide or protein sequences from FASTA files, one-hot encodes them, and stores them in a Dataset.
• FASTQ File Loaders - Reads large sequence datasets from FASTQ files, one-hot encodes them, and stores them in a Dataset.
• Genomic Alignment Parsers - Parses SAM, BAM, CRAM, and FASTA alignment files to extract read-level metadata and sequence information.
• Genomic Pileup Image Generators - Creates multi-channel image representations of genomic read alignments around variant positions for variant calling.
• Fingerprint Generators - Generates fixed-length bit-vector or count-based fingerprints from molecular structures.
• Solubility Estimators - Trains models on molecular structures to estimate compound solubility in solvents.
• Molecule-to-Graph Converters - Converts molecular structures into graph data objects with atom and bond features for GNNs.
• Physicochemical Descriptor Calculators - Computes physicochemical and structural descriptors from molecules for QSAR modeling.
• Binding Pocket Detection - Detects potential binding pockets on protein surfaces using geometric algorithms.
• Conformation Feature Extraction - Extracts descriptive features from protein conformations for downstream modeling tasks.
• Protein Structure Featurizers - Extracts N, CA, and C backbone atom coordinates from PDB files for protein structure modeling.
• Sequence Motif Analyzers - Builds position frequency matrices from multiple sequence alignments for sequence motif analysis.
• Biological Analysis - Democratizes deep learning for drug discovery and chemistry.
• General Tools - Deep learning toolkit for drug discovery and quantum chemistry.

سجل النجوم