/tools

The ESMDiff repository provides a framework for generating protein conformations using structure language models. It employs a discrete variational auto-encoder and conditional language modeling to capture conformational distributions, making it a valuable tool for protein design and molecular dynamics.

S-PLM is a structure-aware protein language model that utilizes contrastive learning to integrate sequence and structural information for generating protein embeddings. It supports downstream tasks such as enzyme classification and protein structure prediction, making it a valuable tool in molecular biology.

Dyna-1 is a model designed to predict micro-millisecond motions in proteins using their sequence and structure. It includes curated datasets for evaluation and is aimed at advancing the understanding of protein dynamics through deep learning techniques.

HuDiff is a tool designed for the humanization of antibodies and nanobodies using diffusion models. It includes training and evaluation scripts, as well as preprocessed datasets for developing humanized antibodies from mouse antibodies.

FLAG is a framework for generating 3D molecules that bind to target proteins using a fragment-based approach. It employs deep generative models to create realistic molecular structures, enhancing the process of structure-based drug design.

GearBind is a pretrainable geometric graph neural network that predicts changes in protein-protein binding affinity. It is pretrained on CATH and fine-tuned on SKEMPI, providing tools for inference and dataset processing related to antibody affinity maturation.

Frame2seq is a tool for generating and scoring protein sequences using a structured-conditioned masked language model. It allows users to design sequences based on PDB structures and provides functionality for scoring these sequences based on their likelihood.

GraDe_IF implements a graph denoising diffusion model for inverse protein folding, allowing for the generation of protein sequences based on structural information. It utilizes advanced machine learning techniques to optimize the design of proteins, making it a valuable tool in molecular biology.

μProtein is a framework that accelerates protein engineering by integrating a deep learning model for predicting mutational effects with a reinforcement learning algorithm for navigating the protein fitness landscape. It aims to enhance the design and optimization of proteins through advanced computational techniques.

The 'timed-design' repository provides a library for protein sequence design using deep learning models. It includes functionalities for predicting protein sequences based on 3D structures and sampling sequences using Monte Carlo methods, making it a valuable tool for molecular biology and computational chemistry.

HalluDesign is an all-atom framework that utilizes a structure prediction model to iteratively co-optimize and co-design protein sequences and structures. It allows for the design of new protein sequences based on structural hallucination, making it a valuable tool in molecular biology and protein engineering.

MegaFold is a system designed to accelerate protein structure prediction models, such as AlphaFold2 and AlphaFold3, by optimizing performance on heterogeneous hardware. It provides enhancements in training speed and memory efficiency, making it a valuable tool for researchers in molecular biology.

Evolla is a generative model designed to decode the molecular language of proteins, facilitating the generation and understanding of protein sequences. It provides tools for protein design and can be utilized for various applications in molecular biology.

BinderFlow is a pipeline for designing protein binders, allowing users to manage and monitor the design process efficiently. It includes features for batch processing, design filtering, and integration with DNA synthesis workflows.

ProteinSolver is a deep neural network tool designed to generate amino acid sequences that can fold into specific protein structures. It utilizes graph neural networks to tackle protein design challenges, making it a valuable resource in the field of bioinformatics.

The METL framework provides tools for pretraining and finetuning biophysics-informed protein language models, enabling users to train models on mutational data and generate predictions. It includes datasets for training and examples for running inference, making it a valuable resource for protein engineering and design.

The 'salad' repository provides a framework for generating protein structures using sparse all-atom denoising models. It allows for various protein design tasks, including unconditional generation and motif scaffolding, leveraging advanced machine learning techniques.

The 'walk-jump' repository provides an implementation of discrete Walk-Jump Sampling (dWJS) for training and sampling in protein design. It includes functionalities for evaluating large molecule descriptors and assessing sample quality, making it a valuable tool in molecular design and optimization.

ConformationalBiasing is a computational tool designed for creating protein variants that favor specific conformational states. It utilizes inverse-folding models to score and predict the effects of mutations on protein function, making it a valuable resource for protein engineering and design.

NA-MPNN is a tool for RNA sequence design and predicting protein-DNA specificity. It provides training and inference code for generating sequences and evaluating specificity, making it relevant for molecular design applications.

FrameDiPT is a tool that utilizes an SE(3) diffusion model for protein structure inpainting and de novo design. It allows for the generation of protein backbones and evaluates their structural properties, making it a valuable resource in molecular biology and computational chemistry.

Equiformer Diffusion is an implementation of denoising diffusion specifically aimed at protein design, utilizing the Equiformer architecture. It incorporates advanced techniques for generating and optimizing protein structures, making it a valuable tool in molecular design.

The 'cgdms' repository provides a Python package for differentiable molecular simulation of proteins using a coarse-grained potential. It includes functionalities for simulating protein dynamics, calculating energies, and designing proteins based on learned potentials.

PackDock is a tool that implements a diffusion-based model for flexible protein-ligand docking, allowing for the effective packing of side chains in protein structures. It integrates with various ligand conformation sampling algorithms to enhance docking results.