/tools

The COMP6 repository contains a benchmark suite for assessing the performance of machine-learning molecular potentials. It includes results and methodologies for evaluating the accuracy of these models in predicting molecular properties.

NeuralMD is a tool designed for simulating protein-ligand binding dynamics using a multi-grained symmetric differential equation model. It includes datasets for training and evaluating models, making it relevant for molecular simulations and property predictions in drug discovery.

The SIGN repository provides an implementation of a structure-aware interactive graph neural network designed to predict protein-ligand binding affinity. It includes preprocessing steps for datasets like PDBbind and CSAR-HiQ, making it a valuable tool for researchers in molecular property prediction.

ProteinReDiff is a framework that utilizes equivariant diffusion-based generative models to redesign ligand-binding proteins. It allows for the generation of high-affinity proteins based on initial sequences and ligand SMILES, facilitating advancements in drug discovery and protein engineering.

InfoAlign is a tool that learns molecular representations by integrating molecular structures, cell morphology, and gene expressions. It provides functionalities for fine-tuning models on various molecular datasets, making it useful for tasks related to molecular property prediction.

moldrug is a Python package that focuses on drug-oriented optimization within the chemical space. It employs a Genetic Algorithm as a search engine and integrates with the CReM library for generating chemical structures.

The Protein-Language-Models repository provides a systematic review of protein language models, covering their architectures, evaluation metrics, and relevant datasets. It also introduces tools for ongoing research in the field of protein modeling and analysis.

ADME@NCATS is an application that hosts prediction models for various ADME properties, utilizing QSAR models to facilitate drug discovery. It allows users to predict molecular properties such as solubility and permeability, making it a valuable resource in computational chemistry.

Madrigal is an open-source model designed to predict the clinical outcomes of drug combinations based on multimodal preclinical data. It provides tools for training and testing models, generating embeddings, and evaluating drug interactions, making it a valuable resource in the field of drug discovery.

Molecular Blender is a Blender plugin that specializes in importing and visualizing molecular data from .xyz files, enabling users to create animations and representations of molecular structures. It supports various visualization styles and can dynamically represent molecular properties such as atomic charges and molecular orbitals.

AlphaFold-StepByStep provides a step-by-step guide to running AlphaFold2 and AlphaFold-Multimer for predicting protein structures. It includes processes for searching homologous sequences, generating unrelaxed models, relaxing them, and sorting by confidence scores.

The Triangle Multiplicative module is a PyTorch implementation that efficiently mixes rows or columns of a 2D feature map. It is utilized in AlphaFold2 for predicting protein structures, contributing to advancements in protein design and molecular modeling.

LLM4Mol is a code implementation designed to leverage large language models for predicting molecular properties. It provides a framework for training and evaluating models specifically aimed at enhancing molecular property prediction capabilities.

ProteinCLIP is a tool that harmonizes protein language models with natural language models to enhance the prediction of protein-protein interactions. It provides pre-trained models and training scripts for generating protein embeddings and classifiers.

MindlessGen is a Python package designed for the semi-automated generation of small molecules using a rule-based algorithm. It allows users to specify element compositions and generates molecules by placing atoms randomly in coordinate space, making it useful for molecular design and exploration.

DENOPTIM is a software package that facilitates the de novo design and virtual screening of functional molecules by assembling building blocks and analyzing their properties. It employs genetic algorithms for optimization and is suitable for various types of chemical entities.

ASE.jl offers Julia bindings for the Atomic Simulation Environment, enabling users to perform molecular simulations and calculations. It facilitates the generation of molecular structures and integrates with the JuLIP framework for enhanced functionality.

The prtm repository provides a deep learning library for protein models, enabling tasks such as protein folding, inverse folding, and ligand docking. It aims to streamline workflows in protein design and structure prediction, making it a valuable tool in computational biology and drug discovery.

This repository contains models for spectral neighbor analysis potential (SNAP) used in molecular simulations. It includes force field parameters and training data, facilitating the development and application of these models in computational chemistry.

The ESM-S repository provides a structure-informed protein language model that enhances the learning of protein representations by integrating structural information without requiring explicit protein structures. It is designed for tasks such as remote homology detection and function prediction in proteins.

Rosetta-DL is a collection of deep-learning packages aimed at predicting and designing biomolecular structures, including proteins and antibodies. It includes tools for protein folding and sequence design, contributing to advancements in molecular biology and computational chemistry.

MolRL-MGPT is a code repository for a NeurIPS 2023 paper that presents a method for de novo drug design using multiple GPT agents in a reinforcement learning framework. It incorporates large molecular datasets and benchmarks for evaluating the generated molecules, making it a relevant tool in the field of molecular design and drug discovery.

flare++ is a software tool that extends the FLARE code to implement Bayesian force fields using sparse Gaussian process regression. It is designed for efficient molecular dynamics simulations and includes features for training force fields based on energy and force data.