/tools

The DeepMind Research repository contains implementations and code for various research projects, including those focused on molecular simulations and protein structure prediction. It aims to accelerate scientific progress by providing tools and datasets for the research community.

AlphaFold is an open-source implementation of a deep learning model that predicts protein structures from amino acid sequences. It utilizes advanced algorithms to provide accurate predictions, significantly aiding in molecular biology research and applications.

AlphaFold 3 is an inference pipeline that allows users to predict the three-dimensional structures of proteins based on their amino acid sequences. It leverages advanced machine learning techniques to provide accurate predictions, which are essential for understanding biomolecular interactions and functions.

The 'esm' repository provides pretrained transformer models for proteins, enabling tasks such as protein design and structure prediction. It includes tools for generating protein structures from sequences and predicting the effects of mutations, making it a valuable resource in molecular biology.

OpenFold is a trainable and memory-efficient PyTorch implementation of AlphaFold 2, focused on predicting protein structures. It allows users to train and infer models for protein folding, contributing to advancements in molecular biology and computational chemistry.

ColabFold is a tool that makes protein folding accessible through Google Colab, utilizing models like AlphaFold for structure prediction. It supports the generation of multiple sequence alignments and facilitates the prediction of protein structures, making it a valuable resource in molecular biology.

RoseTTAFold is a deep learning package that provides models and scripts for predicting protein structures and their interactions. It utilizes a three-track neural network architecture to enhance the accuracy of these predictions, making it a valuable tool in molecular biology and computational chemistry.

Chai-1 is a multi-modal foundation model for predicting the structures of various biomolecules, including proteins, small molecules, DNA, and RNA. It utilizes advanced techniques to achieve state-of-the-art performance in molecular structure prediction across multiple benchmarks.

This repository is an unofficial PyTorch implementation of AlphaFold2, a model that predicts protein structures from amino acid sequences. It utilizes advanced deep learning techniques to generate accurate structural predictions, contributing significantly to the field of molecular biology.

AlphaFold 3 - Pytorch is an implementation of the AlphaFold 3 model from Google DeepMind, designed to predict protein structures using deep learning techniques. It provides functionalities for training and evaluating models on molecular data, particularly focusing on protein structures.

Protenix is an open-source tool designed for high-accuracy biomolecular structure prediction, particularly for proteins. It includes related projects for protein design and benchmarking, enhancing its utility in computational biology and drug discovery.

ProtTrans is a repository that offers pre-trained language models specifically designed for proteins, enabling tasks such as feature extraction, prediction, and protein sequence generation. It supports the bioinformatics community by providing tools for analyzing protein sequences and structures.

PaddleHelix is a bio-computing platform that leverages deep learning for drug discovery, vaccine design, and precision medicine. It offers various applications including molecular property prediction, drug-target interaction prediction, and molecular generation, along with advanced protein structure prediction capabilities.

OpenBioMed is a Python deep learning toolkit designed for AI-empowered biomedicine, offering flexible APIs and over 20 tools for various applications including molecular property prediction, protein folding, and docking. It supports a wide range of molecular types and provides a unified data processing pipeline for handling multi-modal biomedical data.

SimpleFold is a protein folding model that utilizes a generative flow-matching approach to predict protein structures from sequences. It is designed to be efficient and scalable, achieving competitive performance on standard folding benchmarks.

ProteinNet is a standardized dataset designed for machine learning of protein structures, offering sequences, structures, and multiple sequence alignments. It aims to facilitate research in protein structure prediction by providing a consistent framework for training and validation across various methods.

ColabDesign is a tool that makes protein design accessible through Google Colab, utilizing models like TrRosetta and AlphaFold for generating and optimizing protein structures based on sequences. It provides various functionalities for predicting protein structures and sequences, making it a valuable resource in the field of molecular biology.

Open-AF3 is an open-source implementation of AlphaFold 3, a model that predicts the structures of biomolecules such as proteins. It utilizes advanced machine learning techniques to generate accurate structural predictions based on input sequences, making it a valuable tool for researchers in molecular biology.

RoseTTAFold All-Atom is a neural network designed for predicting the structures of various biomolecular assemblies, including proteins, nucleic acids, and small molecules. It provides functionalities for predicting protein-nucleic acid complexes and covalently modified proteins, making it a valuable tool in molecular biology and computational chemistry.

PyRosetta.notebooks offers Jupyter Notebooks that serve as a learning resource for the PyRosetta platform, which is used for biomolecular structure prediction and design. The repository includes tutorials on protein folding, docking, and design, making it a valuable tool for researchers in computational biology and chemistry.

DL4Proteins provides a series of Jupyter notebooks that teach deep learning methodologies for predicting and designing biomolecular structures, particularly proteins. It covers advanced topics such as AlphaFold and graph neural networks, making it a valuable resource for researchers in the field of protein engineering.

FastFold is a high-performance implementation of the Evoformer model used in AlphaFold, aimed at optimizing protein structure prediction on GPU clusters. It enhances training and inference speed while supporting large protein sequences, making it a valuable tool for researchers in molecular biology and computational chemistry.

EquiBind is a geometric deep learning model that predicts the binding location and orientation of small molecules to proteins. It utilizes SE(3)-equivariant neural networks to achieve fast and accurate predictions, making it a valuable tool in drug discovery.

OpenFold3-preview is an open-source biomolecular structure prediction model that aims to replicate the capabilities of AlphaFold3. It supports the prediction of structures for proteins, RNA, and DNA, and includes benchmarking against state-of-the-art models.