/tools

PXDesignBench is a comprehensive evaluation suite for protein design that integrates various state-of-the-art models and standardized pipelines. It supports tasks such as monomer and binder design, allowing for thorough assessment of protein sequences and structures.

dyMEAN is a tool for end-to-end full-atom antibody design, enabling the generation and optimization of antibody structures based on specific epitope definitions. It includes functionalities for complex structure prediction and affinity optimization, making it a valuable resource in drug discovery and protein design.

trRosetta2 provides deep learning models and scripts for predicting protein structures based on inter-residue geometries. It is designed to assist in the molecular design and optimization of protein structures, making it a valuable tool in computational biology.

DrugGPT is a tool that employs a GPT-based strategy to design potential ligands for specific proteins. It utilizes deep learning to explore chemical space and optimize ligand design, enhancing the drug development process.

PPIFlow is a framework for the de novo generation of high-affinity biological binders, including antibodies and nanobodies. It integrates a design workflow that supports various tasks such as binder design, antibody design, and motif scaffolding, utilizing deep learning techniques for protein structure generation.

PLAID is a multimodal generative model designed for generating protein sequences and all-atom structures based on specific prompts. It includes features for both unconditional and conditional sampling, making it a valuable tool for protein design and molecular generation.

FreeBindCraft is a modified version of the BindCraft pipeline that allows for protein design and optimization with an optional bypass of PyRosetta. It incorporates structural relaxation, shape complementarity calculations, and scoring metrics to enhance the design process for peptides and miniproteins.

RareFold is a tool for predicting the structures of single-chain proteins that incorporate rare noncanonical amino acids. It also enables the design of novel peptide binders through its EvoBindRare framework, allowing for flexible and diverse peptide design without prior knowledge of binding sites.

ProGen is an implementation of a language model designed for generating protein sequences, akin to GPT for text. It utilizes deep learning techniques to facilitate the design and generation of proteins, making it a valuable tool in the field of molecular biology.

ESM-GearNet is a codebase for joint representation learning on protein sequences and structures, combining sequence and structure encoders to enhance protein representation. It includes pre-training techniques and is designed for tasks related to protein structure analysis.

FLIP is a collection of tasks designed to evaluate the effectiveness of protein sequence representations in modeling protein design aspects. It includes datasets and benchmarks for assessing machine learning models in the context of protein engineering.

RGN2 is a reference implementation of a recurrent geometric network designed for predicting the 3D structures of orphan or de novo protein sequences. It utilizes deep learning techniques to enhance protein structure prediction, making it a valuable tool in the field of molecular biology.

PepGLAD is a tool for full-atom peptide design that utilizes geometric latent diffusion models to co-design peptide sequences and structures. It supports binding conformation generation and provides datasets for benchmarking the models.

OVO is an open-source ecosystem designed for de novo protein design, integrating models, workflows, and data management. It provides a scalable platform for scaffold and binder design, utilizing advanced methods for protein structure generation and validation.

FlowSite and HarmonicFlow are tools for generating binding structures for single and multi-ligands, as well as designing binding site residues. They utilize advanced generative models to optimize molecular interactions, making them valuable for applications in drug discovery and protein design.

MEAN is a tool for conditional antibody design utilizing a multi-channel equivariant attention network. It provides functionalities for redesigning antibody CDRs and optimizing binding affinities, making it a valuable resource in the field of molecular design and drug discovery.

BAGEL is a customizable Python framework for programmable protein design that formalizes the design task as an optimization over an energy landscape. It includes components for defining energy terms, oracles for model integration, and algorithms for sequence sampling and optimization.

The 'protein-ebm' repository provides a PyTorch implementation of energy-based models aimed at predicting protein conformations at atomic resolution. It includes training code, datasets, and pre-trained model weights, making it a valuable resource for researchers in molecular biology and computational chemistry.

PepFlow is a tool for full-atom peptide design utilizing multi-modal flow matching techniques. It allows for the generation and evaluation of peptides based on their interaction with receptor binding pockets, providing a framework for peptide optimization and design.

ProteinDT is a framework for text-guided protein design that allows for the generation and editing of protein sequences. It utilizes advanced machine learning techniques to optimize protein properties and facilitate design tasks.

ABCFold provides scripts to run AlphaFold3 and other related models for predicting protein structures using multiple sequence alignments and custom templates. It facilitates the generation of protein models and outputs useful visualizations for analysis.

RITA is a family of autoregressive models designed for generating protein sequences. It leverages deep learning techniques to facilitate the design and optimization of proteins, making it a valuable tool in molecular biology and computational chemistry.

ProGen3 is a repository that contains models for scoring and generating protein sequences using machine learning techniques. It allows users to evaluate the likelihood of sequences and generate new protein sequences based on prompts, making it a valuable tool for molecular design in bioinformatics.

PINNACLE is a geometric deep learning model designed to generate contextualized representations of proteins based on their interactions across different cell types and tissues. It aims to improve the understanding of protein functions and therapeutic potentials by incorporating biological context into its modeling approach.