/tools

GenAI4Drug is a survey repository that explores the use of generative AI techniques for de novo drug design, emphasizing the generation of molecules and proteins. It includes discussions on various models, datasets, and metrics relevant to molecular design and property prediction.

InstructPLM is a tool designed for generating protein sequences based on structural instructions using a large protein language model. It allows users to create and evaluate protein designs, achieving state-of-the-art performance on relevant benchmarks.

This repository implements a molecular generative model that utilizes randomized SMILES strings to create and train models for molecule generation. It provides scripts for model creation, training, sampling, and calculating log-likelihoods, facilitating de novo drug design.

DrugGEN is a generative system designed for the de novo creation of drug candidate molecules that are tailored to interact with specific protein targets. It utilizes graph transformer-based generative adversarial networks to generate and evaluate potential drug candidates.

MolEncoder is a Molecular AutoEncoder implemented in PyTorch that allows users to train models on molecular datasets, specifically designed for tasks such as molecular representation and generation. It includes functionalities for downloading datasets and training models, making it a useful tool in computational chemistry and molecular biology.

Dayhoff is a resource that combines extensive protein sequence data with generative language models to predict mutation effects and generate novel protein sequences. It includes datasets and models that facilitate the design and analysis of proteins, making it a valuable tool in molecular biology.

Denovo-Pinal is a tool for designing proteins from natural language descriptions, enabling users to generate novel protein sequences based on specified instructions. It utilizes advanced models to facilitate the design process, making it a valuable resource in the field of molecular biology and protein engineering.

ScoreMD is a framework designed for training energy-based diffusion models that can perform both independent sampling and continuous molecular dynamics simulations. It provides tools for energy estimation and generative modeling in the context of molecular dynamics.

This repository provides a codebase for designing molecules using structured state-space sequence models, enabling users to pre-train and fine-tune models for de novo drug design. It simplifies the process of generating new bioactive molecules with minimal code, making it a valuable tool for researchers in drug discovery.

ShEPhERD is a diffusion generative model designed for bioisosteric drug design, capable of generating new molecules in their 3D conformations based on learned distributions of molecular structures. It includes functionalities for training and inference, making it a valuable tool for molecular design and optimization.

ProteinDJ is a protein design pipeline that integrates multiple software packages to facilitate the design and optimization of protein structures. It allows users to generate new protein sequences and structures through various design modes, leveraging advanced modeling techniques like RFdiffusion and AlphaFold2.

SECSE is a platform for systemic evolutionary chemical space exploration aimed at drug discovery. It utilizes deep learning and fragment-based design to generate novel small molecules, enhancing the hit-finding process in drug development.

ProGen2-finetuning is a tool for finetuning a generative protein language model to generate sequences from selected protein families. It allows users to preprocess data, finetune the model, and sample new protein sequences, making it relevant for protein design and generation tasks.

UniMoMo is a generative modeling tool designed for de novo binder design, enabling the creation of small molecules, peptides, and antibodies. It provides functionalities for training and evaluating models to generate molecular candidates based on specified configurations.

RetroExplainer is a deep-learning framework designed for predicting retrosynthesis pathways with a focus on molecular assembly reasoning and interpretability. It allows users to generate and analyze potential synthetic routes for chemical compounds, making it a useful tool in the field of molecular design.

LigDream is a tool for generating novel molecules based on a reference shape using generative modeling techniques. It utilizes a dataset of drug-like compounds for training and allows for the generation of new compounds through a web interface or locally via Jupyter Notebooks.

PaddleMaterials is an end-to-end toolkit based on the PaddlePaddle deep learning framework, aimed at facilitating the discovery and development of new materials. It supports various tasks such as property prediction and structure generation, making it a valuable resource for researchers in materials science.

RNAFlow is a tool for designing RNA sequences and structures using a flow matching model that incorporates inverse folding techniques. It allows users to generate RNA sequences conditioned on protein structures, facilitating the design of RNA molecules for various applications.

MolDQN-pytorch is a PyTorch implementation of a deep reinforcement learning approach for optimizing molecular properties. It allows users to train models for property optimization tasks, making it a valuable tool in the field of molecular design and drug discovery.

GEMORNA is a deep generative model that designs mRNA sequences with enhanced translational capacity and stability. It supports the generation of coding sequences and untranslated regions, making it a valuable tool for advancing mRNA therapeutics and vaccines.

JAXChem is a JAX-based deep learning library that facilitates complex chemical modeling. It aims to support various chemical modeling tasks, including property prediction and molecular generation.

Architector is a Python package that generates 3D chemical structures of organometallic complexes from minimal 2D information. It supports high-throughput in-silico construction and is capable of producing chemically sensible conformers and stereochemistry for various metal complexes.

ReQFlow is a generative model designed for efficient and high-quality protein backbone generation. It utilizes a rectified quaternion flow approach to enhance the designability and speed of protein structure generation, making it a valuable tool in protein design and molecular biology.

The NVIDIA BioNeMo blueprint provides a framework for generative virtual screening in drug discovery, utilizing advanced AI models to design and optimize small molecules and predict protein-ligand interactions. It integrates various tools for protein structure prediction and molecular generation, facilitating efficient drug discovery workflows.