/tools

MBP is a PyTorch implementation designed for multi-task bioassay pre-training aimed at predicting protein-ligand binding affinities. It utilizes the ChEMBL-Dock dataset, which contains extensive protein-ligand binding data, to train models that can predict binding affinities effectively.

iPPIGAN is a tool for de novo molecular design that utilizes deep molecular generative models to create inhibitors targeting protein-protein interactions. It includes functionalities for training models and generating novel compounds, contributing to drug discovery efforts.

DyScore is an open-source tool that implements a scoring method for identifying true binders and non-binders in drug discovery. It utilizes molecular docking and dynamic feature generation to predict the binding likelihood of compounds to target proteins.

DrugPilot is an advanced LLM-based agent framework that automates and optimizes various aspects of drug discovery. It includes functionalities for predicting drug properties, generating new drug candidates, and classifying drug-target affinities, thereby streamlining the drug discovery process.

The DrugHunting repository provides Python scripts for automating drug discovery processes, including the design and optimization of drug-like molecules. It utilizes stochastic methods and cheminformatics to explore novel chemical spaces, making it suitable for applications like docking and virtual screening.

PoseidonQ is a software tool designed for efficient QSAR modeling, allowing users to build, validate, and deploy predictive models for molecular properties. It supports data extraction from databases and offers various machine learning models for regression and classification tasks in drug discovery.

fpocketR is a command-line tool designed to analyze RNA structures and visualize drug-like RNA-ligand binding pockets. It provides functionalities for characterizing these pockets and supports various analysis workflows for RNA-ligand interactions.

The paccmann_sarscov2 repository provides a pipeline for automating the discovery and synthesis of targeted molecules using machine learning. It includes models for predicting protein-ligand interactions, toxicity, and generative models for both proteins and small molecules.

Dragonfly_gen is a tool for de novo drug design that utilizes deep interactome learning to generate novel molecules based on predefined properties. It allows users to preprocess data, sample from binding sites, and rank generated molecules based on pharmacophore similarity.

This repository provides code and data for predicting protein-ligand binding affinity using deep learning techniques. It includes various encodings for proteins and ligands, and offers datasets for training and testing models in the context of drug discovery.

DMFF-DTA is a dual-modality neural network designed for accurate drug-target affinity prediction by integrating sequence and structural information. It utilizes a binding site-focused graph construction method and provides a processed dataset for training.

The ESPF tool provides a method for generating explainable substructure partition fingerprints for proteins and drugs, facilitating molecular property prediction. It includes datasets and examples for customizing the fingerprint generation process, making it useful for drug discovery and protein design.

MoleculeMO is a tool for multiobjective de novo drug design that utilizes recurrent neural networks to generate and optimize molecules based on various properties. It includes data preprocessing, model training, and validation of generated molecules for their pharmacokinetic properties.

GENiPPI is an interface-aware molecular generative framework aimed at designing modulators for protein-protein interactions. It utilizes a dataset of PPI interfaces to generate novel compounds, enhancing the capabilities of structure-based drug design.

UQ4DD is a Python package designed for uncertainty quantification in drug discovery, specifically for predicting molecular properties and drug-target interactions. It utilizes deep learning techniques to provide estimates of uncertainty alongside property predictions, making it a valuable tool for researchers in the field.

FMol is a simplified drug discovery pipeline that generates SMILE molecular representations using AlphaSMILES, predicts protein structures with AlphaFold, and evaluates druggability using fpocket and Amber. It integrates various molecular modeling techniques to facilitate drug discovery processes.

RELATION is a software tool that implements a deep generative model for structure-based de novo drug design. It allows users to prepare molecular datasets, train models, and sample new molecular structures, making it relevant for drug discovery and molecular design.

PROTON is a graph AI model designed to generate and validate neurological hypotheses across molecular, organoid, and clinical systems. It predicts drug candidates and toxicities related to neurological diseases, leveraging a heterogeneous graph transformer trained on a comprehensive biomedical knowledge graph.

SaLT&PepPr is a language model designed to create peptide-guided protein degraders for targeting undruggable proteins. It utilizes a sequence-based framework to select peptides for therapeutic intervention without requiring structural information.

GPCNDTA is a tool designed for predicting drug-target binding affinity using cross-attention networks enhanced with graph features and pharmacophores. It includes benchmark datasets for training and evaluation, making it suitable for drug discovery applications.

DeepLBVS is a tool for ligand-based virtual screening that utilizes deep learning techniques to predict molecular properties. It generates ECFP fingerprints and performs cross-validation using RandomForest models to assess assay results, making it useful for drug discovery applications.

The Flexible Topology project develops a machine learning-based tool for dynamically designing potential drug molecules. It utilizes PyTorch to create models that predict molecular structures and poses, and incorporates molecular dynamics simulations to optimize these designs toward drug-like candidates.

MolDesigner is an interactive web interface that helps drug developers design efficacious drugs using deep learning predictions. It allows users to input drug molecules and receive real-time predictions on important indices related to drug efficacy, facilitating iterative design.

NHGNN-DTA is a tool designed for predicting drug-target binding affinity using a node-adaptive hybrid graph neural network. It includes functionalities for data preparation, model training, and evaluation, making it relevant for molecular property prediction in the context of drug discovery.