[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"similar-PaddlePaddle--PaddleHelix":3,"tool-PaddlePaddle--PaddleHelix":64},[4,17,27,35,43,56],{"id":5,"name":6,"github_repo":7,"description_zh":8,"stars":9,"difficulty_score":10,"last_commit_at":11,"category_tags":12,"status":16},3808,"stable-diffusion-webui","AUTOMATIC1111\u002Fstable-diffusion-webui","stable-diffusion-webui 是一个基于 Gradio 构建的网页版操作界面，旨在让用户能够轻松地在本地运行和使用强大的 Stable Diffusion 图像生成模型。它解决了原始模型依赖命令行、操作门槛高且功能分散的痛点，将复杂的 AI 绘图流程整合进一个直观易用的图形化平台。\n\n无论是希望快速上手的普通创作者、需要精细控制画面细节的设计师，还是想要深入探索模型潜力的开发者与研究人员，都能从中获益。其核心亮点在于极高的功能丰富度：不仅支持文生图、图生图、局部重绘（Inpainting）和外绘（Outpainting）等基础模式，还独创了注意力机制调整、提示词矩阵、负向提示词以及“高清修复”等高级功能。此外，它内置了 GFPGAN 和 CodeFormer 等人脸修复工具，支持多种神经网络放大算法，并允许用户通过插件系统无限扩展能力。即使是显存有限的设备，stable-diffusion-webui 也提供了相应的优化选项，让高质量的 AI 艺术创作变得触手可及。",162132,3,"2026-04-05T11:01:52",[13,14,15],"开发框架","图像","Agent","ready",{"id":18,"name":19,"github_repo":20,"description_zh":21,"stars":22,"difficulty_score":23,"last_commit_at":24,"category_tags":25,"status":16},1381,"everything-claude-code","affaan-m\u002Feverything-claude-code","everything-claude-code 是一套专为 AI 编程助手（如 Claude Code、Codex、Cursor 等）打造的高性能优化系统。它不仅仅是一组配置文件，而是一个经过长期实战打磨的完整框架，旨在解决 AI 代理在实际开发中面临的效率低下、记忆丢失、安全隐患及缺乏持续学习能力等核心痛点。\n\n通过引入技能模块化、直觉增强、记忆持久化机制以及内置的安全扫描功能，everything-claude-code 能显著提升 AI 在复杂任务中的表现，帮助开发者构建更稳定、更智能的生产级 AI 代理。其独特的“研究优先”开发理念和针对 Token 消耗的优化策略，使得模型响应更快、成本更低，同时有效防御潜在的攻击向量。\n\n这套工具特别适合软件开发者、AI 研究人员以及希望深度定制 AI 工作流的技术团队使用。无论您是在构建大型代码库，还是需要 AI 协助进行安全审计与自动化测试，everything-claude-code 都能提供强大的底层支持。作为一个曾荣获 Anthropic 黑客大奖的开源项目，它融合了多语言支持与丰富的实战钩子（hooks），让 AI 真正成长为懂上",138956,2,"2026-04-05T11:33:21",[13,15,26],"语言模型",{"id":28,"name":29,"github_repo":30,"description_zh":31,"stars":32,"difficulty_score":23,"last_commit_at":33,"category_tags":34,"status":16},2271,"ComfyUI","Comfy-Org\u002FComfyUI","ComfyUI 是一款功能强大且高度模块化的视觉 AI 引擎，专为设计和执行复杂的 Stable Diffusion 图像生成流程而打造。它摒弃了传统的代码编写模式，采用直观的节点式流程图界面，让用户通过连接不同的功能模块即可构建个性化的生成管线。\n\n这一设计巧妙解决了高级 AI 绘图工作流配置复杂、灵活性不足的痛点。用户无需具备编程背景，也能自由组合模型、调整参数并实时预览效果，轻松实现从基础文生图到多步骤高清修复等各类复杂任务。ComfyUI 拥有极佳的兼容性，不仅支持 Windows、macOS 和 Linux 全平台，还广泛适配 NVIDIA、AMD、Intel 及苹果 Silicon 等多种硬件架构，并率先支持 SDXL、Flux、SD3 等前沿模型。\n\n无论是希望深入探索算法潜力的研究人员和开发者，还是追求极致创作自由度的设计师与资深 AI 绘画爱好者，ComfyUI 都能提供强大的支持。其独特的模块化架构允许社区不断扩展新功能，使其成为当前最灵活、生态最丰富的开源扩散模型工具之一，帮助用户将创意高效转化为现实。",107662,"2026-04-03T11:11:01",[13,14,15],{"id":36,"name":37,"github_repo":38,"description_zh":39,"stars":40,"difficulty_score":23,"last_commit_at":41,"category_tags":42,"status":16},3704,"NextChat","ChatGPTNextWeb\u002FNextChat","NextChat 是一款轻量且极速的 AI 助手，旨在为用户提供流畅、跨平台的大模型交互体验。它完美解决了用户在多设备间切换时难以保持对话连续性，以及面对众多 AI 模型不知如何统一管理的痛点。无论是日常办公、学习辅助还是创意激发，NextChat 都能让用户随时随地通过网页、iOS、Android、Windows、MacOS 或 Linux 端无缝接入智能服务。\n\n这款工具非常适合普通用户、学生、职场人士以及需要私有化部署的企业团队使用。对于开发者而言，它也提供了便捷的自托管方案，支持一键部署到 Vercel 或 Zeabur 等平台。\n\nNextChat 的核心亮点在于其广泛的模型兼容性，原生支持 Claude、DeepSeek、GPT-4 及 Gemini Pro 等主流大模型，让用户在一个界面即可自由切换不同 AI 能力。此外，它还率先支持 MCP（Model Context Protocol）协议，增强了上下文处理能力。针对企业用户，NextChat 提供专业版解决方案，具备品牌定制、细粒度权限控制、内部知识库整合及安全审计等功能，满足公司对数据隐私和个性化管理的高标准要求。",87618,"2026-04-05T07:20:52",[13,26],{"id":44,"name":45,"github_repo":46,"description_zh":47,"stars":48,"difficulty_score":23,"last_commit_at":49,"category_tags":50,"status":16},2268,"ML-For-Beginners","microsoft\u002FML-For-Beginners","ML-For-Beginners 是由微软推出的一套系统化机器学习入门课程，旨在帮助零基础用户轻松掌握经典机器学习知识。这套课程将学习路径规划为 12 周，包含 26 节精炼课程和 52 道配套测验，内容涵盖从基础概念到实际应用的完整流程，有效解决了初学者面对庞大知识体系时无从下手、缺乏结构化指导的痛点。\n\n无论是希望转型的开发者、需要补充算法背景的研究人员，还是对人工智能充满好奇的普通爱好者，都能从中受益。课程不仅提供了清晰的理论讲解，还强调动手实践，让用户在循序渐进中建立扎实的技能基础。其独特的亮点在于强大的多语言支持，通过自动化机制提供了包括简体中文在内的 50 多种语言版本，极大地降低了全球不同背景用户的学习门槛。此外，项目采用开源协作模式，社区活跃且内容持续更新，确保学习者能获取前沿且准确的技术资讯。如果你正寻找一条清晰、友好且专业的机器学习入门之路，ML-For-Beginners 将是理想的起点。",84991,"2026-04-05T10:45:23",[14,51,52,53,15,54,26,13,55],"数据工具","视频","插件","其他","音频",{"id":57,"name":58,"github_repo":59,"description_zh":60,"stars":61,"difficulty_score":10,"last_commit_at":62,"category_tags":63,"status":16},3128,"ragflow","infiniflow\u002Fragflow","RAGFlow 是一款领先的开源检索增强生成（RAG）引擎，旨在为大语言模型构建更精准、可靠的上下文层。它巧妙地将前沿的 RAG 技术与智能体（Agent）能力相结合，不仅支持从各类文档中高效提取知识，还能让模型基于这些知识进行逻辑推理和任务执行。\n\n在大模型应用中，幻觉问题和知识滞后是常见痛点。RAGFlow 通过深度解析复杂文档结构（如表格、图表及混合排版），显著提升了信息检索的准确度，从而有效减少模型“胡编乱造”的现象，确保回答既有据可依又具备时效性。其内置的智能体机制更进一步，使系统不仅能回答问题，还能自主规划步骤解决复杂问题。\n\n这款工具特别适合开发者、企业技术团队以及 AI 研究人员使用。无论是希望快速搭建私有知识库问答系统，还是致力于探索大模型在垂直领域落地的创新者，都能从中受益。RAGFlow 提供了可视化的工作流编排界面和灵活的 API 接口，既降低了非算法背景用户的上手门槛，也满足了专业开发者对系统深度定制的需求。作为基于 Apache 2.0 协议开源的项目，它正成为连接通用大模型与行业专有知识之间的重要桥梁。",77062,"2026-04-04T04:44:48",[15,14,13,26,54],{"id":65,"github_repo":66,"name":67,"description_en":68,"description_zh":69,"ai_summary_zh":69,"readme_en":70,"readme_zh":71,"quickstart_zh":72,"use_case_zh":73,"hero_image_url":74,"owner_login":75,"owner_name":75,"owner_avatar_url":76,"owner_bio":77,"owner_company":78,"owner_location":78,"owner_email":78,"owner_twitter":78,"owner_website":79,"owner_url":80,"languages":81,"stars":108,"forks":109,"last_commit_at":110,"license":111,"difficulty_score":112,"env_os":113,"env_gpu":114,"env_ram":115,"env_deps":116,"category_tags":120,"github_topics":121,"view_count":23,"oss_zip_url":78,"oss_zip_packed_at":78,"status":16,"created_at":137,"updated_at":138,"faqs":139,"releases":170},3920,"PaddlePaddle\u002FPaddleHelix","PaddleHelix","Bio-Computing Platform Featuring Large-Scale Representation Learning and Multi-Task Deep Learning “螺旋桨”生物计算工具集","PaddleHelix（螺旋桨）是百度飞桨推出的生物计算开源平台，专注于利用大规模表示学习和多任务深度学习技术，解决生物医药领域中的结构预测与分子设计难题。它能够有效预测蛋白质、核酸及其复合物的三维结构，分析蛋白质与配体的相互作用，并辅助药物分子的优化生成，显著降低了传统实验方法的时间与经济成本。\n\n该平台特别适合生物信息学研究人员、药物研发科学家以及 AI 开发者使用。无论是需要复现前沿算法的学术团队，还是希望将 AI 能力集成到工作流中的工程师，都能从中获益。PaddleHelix 的核心亮点在于其强大的模型家族：例如 HelixFold3 在生物大分子结构预测精度上比肩国际顶尖水平；HelixDock 通过大规模预训练提升了分子对接的准确性；而 HelixFold-Single 则创新性地实现了无需多序列比对即可进行高精度蛋白结构预测。此外，平台不仅提供完整的开源代码与模型参数，还配套了便捷的在线预测服务与 API 接口，让复杂的生物计算任务变得更加触手可及，助力科研创新与药物发现加速落地。","English | [简体中文](README_cn.md)\n\n\u003Cp align=\"center\">\n\u003Cimg src=\".\u002F.github\u002Fpaddlehelix_logo.png\" align=\"middle\" height=\"90%\" width=\"90%\" \u002F>\n\u003C\u002Fp>\n\n------\n[![Version](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frelease\u002FPaddlePaddle\u002FPaddleHelix.svg)](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Freleases)\n![python version](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.6+-orange.svg)\n![support os](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fos-linux%2C%20win%2C%20mac-yellow.svg)\n[![DOI](https:\u002F\u002Fzenodo.org\u002Fbadge\u002F314704349.svg)](https:\u002F\u002Fzenodo.org\u002Fbadge\u002Flatestdoi\u002F314704349)\n\n\n## Latest News\n`2025.07.23` **HelixFold3.2 released!** Compared to HelixFold3, **HelixFold3.2** demonstrates significant improvements in protein-related tasks and structural quality. For implementation details, please see [the code here](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold3).\n\n`2024.11.08` To streamline HelixFold3 integration and support high-throughput use, we introduce a convenient paid API ([usage guide link](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Ftut\u002Fguide\u002Fall\u002Fhelixfold3sdk)) for academic and commercial applications, enabling efficient access to HelixFold3’s structural prediction capabilities.\n\n`2024.08.30` We are excited to announce great news! The initial version of the HelixFold3 server, designed for biomolecular structure prediction, is now available on the PaddleHelix website (https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fall\u002Fhelixfold3\u002Fforecast). We encourage everyone to explore its capabilities and leverage it for impactful and innovative research.\n\n`2024.08.15` PaddleHelix released the codes and model parameters of HelixFold3, biomolecular structure prediction replicating the capabilities of AlphaFold3. HelixFold3 achieves accuracy comparable to AlphaFold3 in predicting the structures of the conventional ligands, nucleic acids, and proteins. The initial release of HelixFold3 is available as open source on GitHub for non-commercial academic research, promising to advance biomolecular research and accelerate discoveries. Refer to [codes](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold3) for more details. \n\n`2024.05.23` PaddleHelix released the codes of HelixDock, a pre-training model on large-scale generated docking conformations to unlock the potential of protein-ligand structure prediction, significantly improving prediction accuracy and generalizability. Please refer to [paper]([https:\u002F\u002Farxiv.org\u002Fabs\u002F2310.13913) and [codes](.\u002Fapps\u002Fmolecular_docking\u002Fhelixdock) for more details. Welcome to [PaddleHelix website](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fhelix-dock\u002Fforecast) to try out the structure prediction online service. \n\n`2024.05.13` Paper \"Multi-purpose RNA Language Modeling with Motif-aware Pre-training and Type-guided Fine-tuning\" is accepted by Nature Machine Intelligence. Please refer to [paper](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs42256-024-00836-4) and [codes](https:\u002F\u002Fgithub.com\u002FCatIIIIIIII\u002FRNAErnie) for more details.\n\n\n`2024.04.16` PaddleHelix released the technical report of HelixFold-Multimer, a protein complex structure prediction model which achieves remarkable success in antigen-antibody and peptide-protein structure prediction. Please refer to the [report](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.10260v2) for more details. The online structure prediction services for general and antigen-antibody protein complex are now available at [link1](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fprotein-complex\u002Fforecast) and [link2](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002FKYKT\u002Fforecast) on the PaddleHelix platform respectively.\n\n`2023.10.09` The work of HelixFold-Single titled with \"A method for multiple-sequence-alignment-free protein structure prediction using a protein language model\" is accepted by Nature Machine Intelligence. Please refer to [paper](https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs42256-023-00721-6) for more details.\n\n\n`2022.12.08` Paper \"HelixMO: Sample-Efficient Molecular Optimization in Scene-Sensitive Latent Space\" is accepted by **BIBM 2022**. Please refere to [link1](https:\u002F\u002Fwww.computer.org\u002Fcsdl\u002Fproceedings-article\u002Fbibm\u002F2022\u002F09995561\u002F1JC23yWxizC) or [link2](https:\u002F\u002Faps.arxiv.org\u002Fabs\u002F2112.00905) for more details. We also deployed the drug design service on the website [PaddleHelix](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fdrugdesign\u002Fforecast).\n\n`2022.08.11` PaddleHelix released the codes of HelixGEM-2, a novel Molecular Property Prediction Network that models full-range many-body interactions. And it ranked 1st in the OGB [PCQM4Mv2](https:\u002F\u002Fogb.stanford.edu\u002Fdocs\u002Flsc\u002Fleaderboards\u002F) leaderboard. Please refer to [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2208.05863) and [codes](.\u002Fapps\u002Fpretrained_compound\u002FChemRL\u002FGEM-2) for more details.\n\n`2022.07.29` PaddleHelix released the codes of HelixFold-Single, an **MSA-free** protein structure prediction pipeline relying on only the primary sequences, which can **predict the protein structures within seconds**. Please refer to [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.13921) and [codes](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold-single) for more details. Welcome to [PaddleHelix website](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fprotein-single\u002Fforecast) to try out the structure prediction online service.\n\n`2022.07.18` PaddleHelix fully released HelixFold including training and inference pipeline. **The complete training time are optimized from 11 days to 5.12 days. Ultra-long monomer protein (around 6600 AA) prediction is supported now**. Please refer to [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.05477) and [codes](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold) for more details.\n\n`2022.07.07` Paper \"BatchDTA: implicit batch alignment enhances deep learning-based drug–target affinity estimation\" is published in **Briefings in Bioinformatics**. Please refer to [paper](https:\u002F\u002Facademic.oup.com\u002Fbib\u002Fadvance-article-abstract\u002Fdoi\u002F10.1093\u002Fbib\u002Fbbac260\u002F6632927) and [codes](.\u002Fapps\u002Fdrug_target_interaction\u002Fbatchdta) for more details.\n\n`2022.05.24` Paper \"HelixADMET: a robust and endpoint extensible ADMET system incorporating self-supervised knowledge transfer\" is published in **Bioinformatics**. Refer to [paper](https:\u002F\u002Facademic.oup.com\u002Fbioinformatics\u002Fadvance-article-abstract\u002Fdoi\u002F10.1093\u002Fbioinformatics\u002Fbtac342\u002F6590643) for more information.\n\n`2022.02.07` Paper \"Geometry-enhanced molecular representation learning for property prediction\" is published in **Nature Machine Intelligence**. Please refer to [paper](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs42256-021-00438-4) and [codes](.\u002Fapps\u002Fpretrained_compound\u002FChemRL\u002FGEM) to explore the algorithm.\n\n\u003Cdetails>\n\u003Csummary>More news ...\u003C\u002Fsummary>\n\n`2022.01.07` PaddleHelix released the reproduction of [AlphaFold 2](https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41586-021-03819-2) inference pipeline using PaddlePaddle in [HelixFold](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold).\n\n`2021.11.23` Paper \"Multimodal Pre-Training Model for Sequence-based Prediction of Protein-Protein Interaction\" is accepted by [MLCB 2021](https:\u002F\u002Fsites.google.com\u002Fcs.washington.edu\u002Fmlcb2021\u002Fhome). Please refer to [paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2112.04814) and [code](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Ftree\u002Fdev\u002Fapps\u002Fprotein_protein_interaction) for more details.\n\n`2021.10.25` Paper \"Docking-based Virtual Screening with Multi-Task Learning\" is accepted by [BIBM 2021](https:\u002F\u002Fieeebibm.org\u002FBIBM2021\u002F).\n\n`2021.09.29` Paper \"Property-Aware Relation Networks for Few-shot Molecular Property Prediction\" is accepted by [NeurIPS 2021](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F91bc333f6967019ac47b49ca0f2fa757-Abstract.html) as a Spotlight Paper. Please refer to [PAR](.\u002Fapps\u002Ffewshot_molecular_property) for more details.\n\n`2021.07.29` PaddleHelix released a novel geometry-level molecular pre-training model, taking advantage of the 3D spatial structures of the molecules. Please refer to [GEM](.\u002Fapps\u002Fpretrained_compound\u002FChemRL\u002FGEM) for more details.\n\n`2021.06.17` PaddleHelix team won the 2nd place in the [OGB-LCS KDD Cup 2021 PCQM4M-LSC track](https:\u002F\u002Fogb.stanford.edu\u002Fkddcup2021\u002Fresults\u002F), predicting DFT-calculated HOMO-LUMO energy gap of molecules. Please refer to [the solution](.\u002Fcompetition\u002Fkddcup2021-PCQM4M-LSC) for more details.\n\n`2021.05.20` PaddleHelix v1.0 released. 1) Update from static framework to dynamic framework; 2) Add new applications: molecular generation and drug-drug synergy.\n\n`2021.05.18` Paper \"Structure-aware Interactive Graph Neural Networks for the Prediction of Protein-Ligand Binding Affinity\" is accepted by [KDD 2021](https:\u002F\u002Fkdd.org\u002Fkdd2021\u002Faccepted-papers\u002Findex). The code is available at [here](.\u002Fapps\u002Fdrug_target_interaction\u002Fsign).\n\n`2021.03.15` PaddleHelix team ranks 1st in the ogbg-molhiv and ogbg-molpcba of [OGB](https:\u002F\u002Fogb.stanford.edu\u002Fdocs\u002Fleader_graphprop\u002F), predicting the molecular properties.\n\u003C\u002Fdetails>\n\n------\n\n## Introduction\nPaddleHelix is a bio-computing tool, taking advantage of the machine learning approaches, especially deep neural networks, for facilitating the development of the following areas:\n* **Drug Discovery**. Provide 1) Large-scale pre-training models: compounds and proteins; 2) Various applications: molecular property prediction, drug-target affinity prediction, and molecular generation.\n* **Vaccine Design**. Provide RNA design algorithms, including LinearFold and LinearPartition.\n* **Precision Medicine**. Provide application of drug-drug synergy.\n\n\u003Cp align=\"center\">\n\u003Cimg src=\".github\u002FPaddleHelix_Structure.png\" align=\"middle\" heigh=\"70%\" width=\"70%\" \u002F>\n\u003C\u002Fp>\n\n## Resources\n### Application Platform\n**[PaddleHelix platform](https:\u002F\u002Fpaddlehelix.baidu.com\u002F)** provides the AI + biochemistry abilities for the scenarios of drug discovery, vaccine design and precision medicine.\n\n### Installation Guide\nPaddleHelix is a bio-computing repository based on [PaddlePaddle](https:\u002F\u002Fgithub.com\u002Fpaddlepaddle\u002Fpaddle), a high-performance Parallelized Deep Learning Platform. The installation prerequisites and guide can be found [here](.\u002Finstallation_guide.md).\n\n### Tutorials\nWe provide abundant [tutorials](.\u002Ftutorials) to help you navigate the repository and start quickly.\n* **Drug Discovery**\n  - [Compound Representation Learning and Property Prediction](.\u002Ftutorials\u002Fcompound_property_prediction_tutorial.ipynb)\n  - [Protein Representation Learning and Property Prediction](.\u002Ftutorials\u002Fprotein_pretrain_and_property_prediction_tutorial.ipynb)\n  - [Predicting Drug-Target Interaction: GraphDTA](.\u002Ftutorials\u002Fdrug_target_interaction_graphdta_tutorial.ipynb), [MolTrans](.\u002Ftutorials\u002Fdrug_target_interaction_moltrans_tutorial.ipynb)\n  - [Molecular Generation](.\u002Ftutorials\u002Fmolecular_generation_tutorial.ipynb)\n* **Vaccine Design**\n  - [Predicting RNA Secondary Structure](.\u002Ftutorials\u002Flinearrna_tutorial.ipynb)\n\n### Examples\nWe also provide [examples](.\u002Fapps) that implement various algorithms and show the methods running the algorithms:\n* **Pretraining**\n  - [Representation Learning - Compounds](.\u002Fapps\u002Fpretrained_compound)\n  - [Representation Learning - Proteins](.\u002Fapps\u002Fpretrained_protein)\n* **Drug Discovery and Precision Medicine**\n  - [Drug-Target Interaction](.\u002Fapps\u002Fdrug_target_interaction)\n  - [Molecular Generation](.\u002Fapps\u002Fmolecular_generation)\n  - [Drug Drug Synergy](.\u002Fapps\u002Fdrug_drug_synergy)\n  - [Few-shot Molecular Property Prediction](.\u002Fapps\u002Ffewshot_molecular_property)\n* **Vaccine Design**\n  - [LinearRNA](.\u002Fc\u002Fpahelix\u002Ftoolkit\u002Flinear_rna)\n* **Protein Structure Prediction**\n  - [HelixFold](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold)\n  - [HelixFold-Single](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold-single)\n  - [HelixFold3](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold3)\n\n### Competition Solutions\nThe PaddleHelix team participated in multiple competitions related to bio-computing. The solutions can be found [here](.\u002Fcompetition).\n\n### Guide for Developers\n* To develop new functions based on the source code of PaddleHelix, please refer to [guide for developers](.\u002Fdeveloper_guide.md).\n* For more details on the APIs, please refer to the [documents](https:\u002F\u002Fpaddlehelix.readthedocs.io\u002Fen\u002Fdev\u002F).\n\n------\n\n## Copyright and License\nShield: [![CC BY-NC-SA 4.0][cc-by-nc-sa-shield]][cc-by-nc-sa]\n\nThis work is licensed under a\n[Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License][cc-by-nc-sa].\n\n[![CC BY-NC-SA 4.0][cc-by-nc-sa-image]][cc-by-nc-sa]\n\n[cc-by-nc-sa]: http:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby-nc-sa\u002F4.0\u002F\n[cc-by-nc-sa-image]: https:\u002F\u002Flicensebuttons.net\u002Fl\u002Fby-nc-sa\u002F4.0\u002F88x31.png\n[cc-by-nc-sa-shield]: https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY--NC--SA%204.0-lightgrey.svg\n","英语 | [简体中文](README_cn.md)\n\n\u003Cp align=\"center\">\n\u003Cimg src=\".\u002F.github\u002Fpaddlehelix_logo.png\" align=\"middle\" height=\"90%\" width=\"90%\" \u002F>\n\u003C\u002Fp>\n\n------\n[![版本](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Frelease\u002FPaddlePaddle\u002FPaddleHelix.svg)](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Freleases)\n![python版本](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fpython-3.6+-orange.svg)\n![支持操作系统](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Fos-linux%2C%20win%2C%20mac-yellow.svg)\n[![DOI](https:\u002F\u002Fzenodo.org\u002Fbadge\u002F314704349.svg)](https:\u002F\u002Fzenodo.org\u002Fbadge\u002Flatestdoi\u002F314704349)\n\n\n## 最新消息\n`2025.07.23` **HelixFold3.2发布！** 相较于HelixFold3，**HelixFold3.2** 在蛋白质相关任务和结构质量方面均有显著提升。具体实现细节请参阅[此处代码](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold3)。\n\n`2024.11.08` 为简化HelixFold3的集成并支持高通量应用，我们推出了一款便捷的付费API（[使用指南链接](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Ftut\u002Fguide\u002Fall\u002Fhelixfold3sdk)），适用于学术及商业场景，可高效调用HelixFold3的结构预测能力。\n\n`2024.08.30` 我们很高兴地宣布一个好消息！专为生物分子结构预测设计的HelixFold3服务器初版现已在PaddleHelix官网上线（https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fall\u002Fhelixfold3\u002Fforecast）。欢迎大家探索其功能，并将其应用于具有影响力和创新性的研究中。\n\n`2024.08.15` PaddleHelix发布了HelixFold3的代码及模型参数，该模型能够复现AlphaFold3的能力，用于生物分子结构预测。HelixFold3在常规配体、核酸和蛋白质结构预测方面的准确性与AlphaFold3相当。HelixFold3的初始版本已在GitHub上以开源形式发布，供非商业性学术研究使用，有望推动生物分子研究并加速科学发现。更多详情请参阅[代码](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold3)。\n\n`2024.05.23` PaddleHelix发布了HelixDock的代码，这是一种基于大规模生成对接构象进行预训练的模型，旨在释放蛋白质-配体结构预测的潜力，显著提升预测精度和泛化能力。更多详情请参阅[论文]([https:\u002F\u002Farxiv.org\u002Fabs\u002F2310.13913) 和 [代码](.\u002Fapps\u002Fmolecular_docking\u002Fhelixdock)。欢迎访问[PaddleHelix官网](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fhelix-dock\u002Fforecast)，体验在线结构预测服务。\n\n`2024.05.13` 论文《基于基序感知预训练与类型引导微调的多用途RNA语言建模》已被Nature Machine Intelligence接收。更多详情请参阅[论文](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs42256-024-00836-4)和[代码](https:\u002F\u002Fgithub.com\u002FCatIIIIIIII\u002FRNAErnie)。\n\n\n`2024.04.16` PaddleHelix发布了HelixFold-Multimer的技术报告，该模型用于蛋白质复合物结构预测，在抗原-抗体及肽-蛋白质结构预测方面取得了显著成果。更多详情请参阅[报告](https:\u002F\u002Farxiv.org\u002Fabs\u002F2404.10260v2)。目前，PaddleHelix平台已分别提供通用型和抗原-抗体蛋白质复合物的在线结构预测服务，访问链接分别为[link1](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fprotein-complex\u002Fforecast)和[link2](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002FKYKT\u002Fforecast)。\n\n`2023.10.09` 题为“利用蛋白质语言模型进行无需多序列比对的蛋白质结构预测方法”的HelixFold-Single相关工作已被Nature Machine Intelligence接收。更多详情请参阅[论文](https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs42256-023-00721-6)。\n\n\n`2022.12.08` 论文《HelixMO：场景敏感潜在空间中的样本高效分子优化》已被**BIBM 2022** 接收。更多详情请参阅[link1](https:\u002F\u002Fwww.computer.org\u002Fcsdl\u002Fproceedings-article\u002Fbibm\u002F2022\u002F09995561\u002F1JC23yWxizC) 或 [link2](https:\u002F\u002Faps.arxiv.org\u002Fabs\u002F2112.00905)。同时，我们还在[PaddleHelix官网](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fdrugdesign\u002Fforecast)上线了药物设计服务。\n\n`2022.08.11` PaddleHelix发布了HelixGEM-2的代码，这是一种新型分子性质预测网络，能够建模全范围的多体相互作用。该模型在OGB [PCQM4Mv2](https:\u002F\u002Fogb.stanford.edu\u002Fdocs\u002Flsc\u002Fleaderboards\u002F)排行榜上位居第一。更多详情请参阅[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2208.05863)和[代码](.\u002Fapps\u002Fpretrained_compound\u002FChemRL\u002FGEM-2)。\n\n`2022.07.29` PaddleHelix发布了HelixFold-Single的代码，这是一个仅依赖于一级序列的**无MSA**蛋白质结构预测流程，能够在**几秒钟内预测蛋白质结构**。更多详情请参阅[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.13921)和[代码](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold-single)。欢迎访问[PaddleHelix官网](https:\u002F\u002Fpaddlehelix.baidu.com\u002Fapp\u002Fdrug\u002Fprotein-single\u002Fforecast)，体验在线结构预测服务。\n\n`2022.07.18` PaddleHelix全面发布了包括训练与推理流程在内的HelixFold。**完整训练时间已从11天优化至5.12天。现已支持超长单体蛋白质（约6600个氨基酸）的预测**。更多详情请参阅[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2207.05477)和[代码](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold)。\n\n`2022.07.07` 论文《BatchDTA：隐式批处理对齐增强基于深度学习的药物-靶标亲和力估计》发表于**Briefings in Bioinformatics**。更多详情请参阅[论文](https:\u002F\u002Facademic.oup.com\u002Fbib\u002Fadvance-article-abstract\u002Fdoi\u002F10.1093\u002Fbib\u002Fbbac260\u002F6632927)和[代码](.\u002Fapps\u002Fdrug_target_interaction\u002Fbatchdta)。\n\n`2022.05.24` 论文《HelixADMET：一种结合自监督知识迁移的稳健且可扩展终点的ADMET系统》发表于**Bioinformatics**。更多信息请参阅[论文](https:\u002F\u002Facademic.oup.com\u002Fbioinformatics\u002Fadvance-article-abstract\u002Fdoi\u002F10.1093\u002Fbioinformatics\u002Fbtac342\u002F6590643)。\n\n`2022.02.07` 论文《面向性质预测的几何增强分子表征学习》发表于**Nature Machine Intelligence**。更多详情请参阅[论文](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs42256-021-00438-4)和[代码](.\u002Fapps\u002Fpretrained_compound\u002FChemRL\u002FGEM)，以深入了解该算法。\n\n\u003Cdetails>\n\u003Csummary>更多新闻...\u003C\u002Fsummary>\n\n`2022.01.07` PaddleHelix发布了基于PaddlePaddle的[AlphaFold 2](https:\u002F\u002Fdoi.org\u002F10.1038\u002Fs41586-021-03819-2)推理流程的复现版本，即[HelixFold](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold)。\n\n`2021年11月23日` 论文《基于序列的蛋白质-蛋白质相互作用预测的多模态预训练模型》被[MLCB 2021](https:\u002F\u002Fsites.google.com\u002Fcs.washington.edu\u002Fmlcb2021\u002Fhome)接收。更多详情请参阅[论文](https:\u002F\u002Farxiv.org\u002Fabs\u002F2112.04814)和[代码](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Ftree\u002Fdev\u002Fapps\u002Fprotein_protein_interaction)。\n\n`2021年10月25日` 论文《基于对接的多任务学习虚拟筛选》被[BIBM 2021](https:\u002F\u002Fieeebibm.org\u002FBIBM2021\u002F)接收。\n\n`2021年9月29日` 论文《面向少样本分子性质预测的属性感知关系网络》作为Spotlight Paper被[NeurIPS 2021](https:\u002F\u002Fpapers.nips.cc\u002Fpaper\u002F2021\u002Fhash\u002F91bc333f6967019ac47b49ca0f2fa757-Abstract.html)接收。更多详情请参阅[PAR](.\u002Fapps\u002Ffewshot_molecular_property)。\n\n`2021年7月29日` PaddleHelix发布了一种新颖的几何级分子预训练模型，该模型充分利用了分子的三维空间结构。更多详情请参阅[GEM](.\u002Fapps\u002Fpretrained_compound\u002FChemRL\u002FGEM)。\n\n`2021年6月17日` PaddleHelix团队在[OGB-LCS KDD Cup 2021 PCQM4M-LSC赛道](https:\u002F\u002Fogb.stanford.edu\u002Fkddcup2021\u002Fresults\u002F)中获得第二名，任务是预测分子的DFT计算HOMO-LUMO能隙。更多详情请参阅[解决方案](.\u002Fcompetition\u002Fkddcup2021-PCQM4M-LSC)。\n\n`2021年5月20日` PaddleHelix v1.0版本发布。更新内容包括：1) 从静态框架升级为动态框架；2) 新增应用：分子生成和药物协同作用预测。\n\n`2021年5月18日` 论文《用于预测蛋白质-配体结合亲和力的结构感知交互式图神经网络》被[KDD 2021](https:\u002F\u002Fkdd.org\u002Fkdd2021\u002Faccepted-papers\u002Findex)接收。代码可在[这里](.\u002Fapps\u002Fdrug_target_interaction\u002Fsign)获取。\n\n`2021年3月15日` PaddleHelix团队在[OGB](https:\u002F\u002Fogb.stanford.edu\u002Fdocs\u002Fleader_graphprop\u002F)的ogbg-molhiv和ogbg-molpcba任务中排名第一，任务是预测分子性质。\n\u003C\u002Fdetails>\n\n------\n\n\n\n## 简介\nPaddleHelix是一款生物计算工具，利用机器学习方法，尤其是深度神经网络，助力以下领域的发展：\n* **药物研发**。提供1) 大规模预训练模型：化合物与蛋白质；2) 多种应用：分子性质预测、药物-靶点亲和力预测以及分子生成。\n* **疫苗设计**。提供RNA设计算法，包括LinearFold和LinearPartition。\n* **精准医疗**。提供药物协同作用预测应用。\n\n\u003Cp align=\"center\">\n\u003Cimg src=\".github\u002FPaddleHelix_Structure.png\" align=\"middle\" height=\"70%\" width=\"70%\" \u002F>\n\u003C\u002Fp>\n\n## 资源\n### 应用平台\n**[PaddleHelix平台](https:\u002F\u002Fpaddlehelix.baidu.com\u002F)**为药物研发、疫苗设计和精准医疗场景提供AI+生物化学能力。\n\n### 安装指南\nPaddleHelix是一个基于[PaddlePaddle](https:\u002F\u002Fgithub.com\u002Fpaddlepaddle\u002Fpaddle)的生物计算库，PaddlePaddle是一个高性能的并行化深度学习平台。安装前提条件及指南请参阅[此处](.\u002Finstallation_guide.md)。\n\n### 教程\n我们提供了丰富的[教程](.\u002Ftutorials)，帮助您快速上手并熟悉本仓库。\n* **药物研发**\n  - [化合物表示学习与性质预测](.\u002Ftutorials\u002Fcompound_property_prediction_tutorial.ipynb)\n  - [蛋白质表示学习与性质预测](.\u002Ftutorials\u002Fprotein_pretrain_and_property_prediction_tutorial.ipynb)\n  - [药物-靶点相互作用预测：GraphDTA](.\u002Ftutorials\u002Fdrug_target_interaction_graphdta_tutorial.ipynb)、[MolTrans](.\u002Ftutorials\u002Fdrug_target_interaction_moltrans_tutorial.ipynb)\n  - [分子生成](.\u002Ftutorials\u002Fmolecular_generation_tutorial.ipynb)\n* **疫苗设计**\n  - [RNA二级结构预测](.\u002Ftutorials\u002Flinearrna_tutorial.ipynb)\n\n### 示例\n我们还提供了[示例](.\u002Fapps)，实现了多种算法，并演示了这些算法的具体运行方式：\n* **预训练**\n  - [化合物表示学习](.\u002Fapps\u002Fpretrained_compound)\n  - [蛋白质表示学习](.\u002Fapps\u002Fpretrained_protein)\n* **药物研发与精准医疗**\n  - [药物-靶点相互作用](.\u002Fapps\u002Fdrug_target_interaction)\n  - [分子生成](.\u002Fapps\u002Fmolecular_generation)\n  - [药物协同作用](.\u002Fapps\u002Fdrug_drug_synergy)\n  - [少样本分子性质预测](.\u002Fapps\u002Ffewshot_molecular_property)\n* **疫苗设计**\n  - [LinearRNA](.\u002Fc\u002Fpahelix\u002Ftoolkit\u002Flinear_rna)\n* **蛋白质结构预测**\n  - [HelixFold](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold)\n  - [HelixFold-Single](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold-single)\n  - [HelixFold3](.\u002Fapps\u002Fprotein_folding\u002Fhelixfold3)\n\n### 竞赛解决方案\nPaddleHelix团队参与了多项生物计算相关的竞赛，解决方案可在此处查看[竞赛解决方案](.\u002Fcompetition)。\n\n### 开发者指南\n* 如需基于PaddleHelix源码开发新功能，请参阅[开发者指南](.\u002Fdeveloper_guide.md)。\n* 更多API细节请参阅[文档](https:\u002F\u002Fpaddlehelix.readthedocs.io\u002Fen\u002Fdev\u002F)。\n\n------\n\n## 版权与许可\n盾牌：[![CC BY-NC-SA 4.0][cc-by-nc-sa-shield]][cc-by-nc-sa]\n\n本作品采用[知识共享署名-非商业性使用-相同方式共享4.0国际许可协议][cc-by-nc-sa]授权。\n\n[![CC BY-NC-SA 4.0][cc-by-nc-sa-image]][cc-by-nc-sa]\n\n[cc-by-nc-sa]: http:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby-nc-sa\u002F4.0\u002F\n[cc-by-nc-sa-image]: https:\u002F\u002Flicensebuttons.net\u002Fl\u002Fby-nc-sa\u002F4.0\u002F88x31.png\n[cc-by-nc-sa-shield]: https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC%20BY--NC--SA%204.0-lightgrey.svg","# PaddleHelix 快速上手指南\n\nPaddleHelix 是基于百度飞桨（PaddlePaddle）开发的生物计算工具集，利用深度学习技术赋能药物发现、疫苗设计和精准医疗。本指南将帮助您快速完成环境配置并运行基础示例。\n\n## 1. 环境准备\n\n在开始之前，请确保您的开发环境满足以下要求：\n\n*   **操作系统**：Linux, Windows, macOS\n*   **Python 版本**：3.6 及以上 (推荐 3.8+)\n*   **核心依赖**：[PaddlePaddle](https:\u002F\u002Fwww.paddlepaddle.org.cn\u002F) (飞桨深度学习框架)\n*   **硬件建议**：如需运行大型蛋白结构预测模型（如 HelixFold3），建议使用配备 NVIDIA GPU 的环境并安装 CUDA 版本的 PaddlePaddle。\n\n> **国内加速提示**：建议配置 pip 使用清华或阿里镜像源以提升下载速度。\n> ```bash\n> pip config set global.index-url https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n> ```\n\n## 2. 安装步骤\n\n### 第一步：安装 PaddlePaddle\n根据您的硬件环境选择安装 CPU 或 GPU 版本。\n\n**CPU 版本：**\n```bash\npython -m pip install paddlepaddle -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n\n**GPU 版本 (以 CUDA 11.2 为例)：**\n```bash\npython -m pip install paddlepaddle-gpu==2.6.0 -i https:\u002F\u002Fpypi.tuna.tsinghua.edu.cn\u002Fsimple\n```\n*(注：具体 GPU 版本请参考 [PaddlePaddle 官网安装指引](https:\u002F\u002Fwww.paddlepaddle.org.cn\u002Finstall\u002Fquick) 匹配您的 CUDA\u002FcuDNN 版本)*\n\n### 第二步：获取 PaddleHelix 源码\n通过 Git 克隆项目仓库：\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix.git\ncd PaddleHelix\n```\n\n### 第三步：安装项目依赖\n进入项目目录后，安装所需的 Python 依赖包：\n```bash\npip install -r requirements.txt\n```\n*(如果 `requirements.txt` 不存在或需安装特定子模块依赖，请参考对应 `apps` 目录下的说明)*\n\n## 3. 基本使用\n\nPaddleHelix 提供了丰富的教程（Tutorials）和示例（Examples）。以下以**化合物性质预测**为例，展示如何快速运行一个基础任务。\n\n### 运行 Jupyter Notebook 教程\n项目 `tutorials` 目录下包含了交互式教程。确保已安装 `jupyter`：\n```bash\npip install jupyter\n```\n\n启动 Jupyter Notebook 并打开化合物性质预测教程：\n```bash\njupyter notebook tutorials\u002Fcompound_property_prediction_tutorial.ipynb\n```\n在 Notebook 中按顺序执行单元格，即可完成从数据加载、模型构建到训练预测的全流程。\n\n### 运行命令行示例 (以 HelixFold-Single 为例)\n如果您想快速体验蛋白结构预测（无需多序列比对 MSA），可以使用 `HelixFold-Single`。\n\n1.  进入对应应用目录：\n    ```bash\n    cd apps\u002Fprotein_folding\u002Fhelixfold-single\n    ```\n2.  查看该目录下的 `README.md` 或脚本，通常运行推理的命令格式如下（需准备输入序列文件 `input.fasta`）：\n    ```bash\n    python run_inference.py --input_fasta=input.fasta --output_dir=output\n    ```\n\n### 更多资源\n*   **在线体验**：访问 [PaddleHelix 开放平台](https:\u002F\u002Fpaddlehelix.baidu.com\u002F) 直接使用云端算力进行药物发现和蛋白结构预测，无需本地部署。\n*   **详细文档**：更多算法细节和 API 文档请参阅 [官方文档](https:\u002F\u002Fpaddlehelix.readthedocs.io\u002Fen\u002Fdev\u002F)。\n*   **模型列表**：支持的任务包括小分子生成、药物靶点亲和力预测、RNA 二级结构预测等，详见 `.\u002Fapps` 目录。","某生物医药公司的算法团队正致力于针对一种新型病毒靶点开发小分子抑制剂，急需快速预测药物分子与蛋白质靶点的结合模式以筛选候选化合物。\n\n### 没有 PaddleHelix 时\n- **预测精度受限**：传统对接软件难以准确模拟蛋白质 - 配体复合物在真实生理环境下的动态构象，导致虚拟筛选的假阳性率极高。\n- **多模态支持缺失**：面对涉及核酸、蛋白质及小分子配体的复杂相互作用，现有工具往往只能处理单一类型，无法进行联合结构预测。\n- **研发周期漫长**：依赖昂贵的商业软件或低效的开源方案，完成一次高精度的全原子结构预测需数天时间，严重拖慢药物发现进程。\n- **模型复现困难**：缺乏类似 AlphaFold3 能力的开源替代方案，团队难以在本地部署并进行针对特定抗原 - 抗体复合物的定制化微调。\n\n### 使用 PaddleHelix 后\n- **精度显著提升**：利用 HelixDock 和 HelixFold3 模型，团队成功复现了接近 AlphaFold3 的预测精度，大幅提高了对常规配体及核酸结合位点的识别准确率。\n- **全场景覆盖**：PaddleHelix 原生支持蛋白质、核酸与小分子的多任务深度学习，轻松应对抗原 - 抗体复合物及复杂药物分子的联合结构建模。\n- **效率飞跃式增长**：借助高吞吐量的 API 服务和本地代码库，将原本需要数天的结构预测任务缩短至小时级，加速了先导化合物的优化迭代。\n- **灵活定制研发**：基于开源的 HelixFold-Multimer 等模块，团队可针对特定病毒变异株快速微调模型，低成本构建专属的生物计算流水线。\n\nPaddleHelix 通过提供大规模表示学习与多任务深度学习能力，将生物大分子结构预测从“高门槛实验”转变为“高效能计算”，极大缩短了创新药物的研发窗口期。","https:\u002F\u002Foss.gittoolsai.com\u002Fimages\u002FPaddlePaddle_PaddleHelix_e1114fb9.png","PaddlePaddle","https:\u002F\u002Foss.gittoolsai.com\u002Favatars\u002FPaddlePaddle_0457ef24.jpg","",null,"http:\u002F\u002Fpaddlepaddle.org","https:\u002F\u002Fgithub.com\u002FPaddlePaddle",[82,86,90,94,98,102,105],{"name":83,"color":84,"percentage":85},"Python","#3572A5",82.2,{"name":87,"color":88,"percentage":89},"Jupyter Notebook","#DA5B0B",10.2,{"name":91,"color":92,"percentage":93},"C++","#f34b7d",6.6,{"name":95,"color":96,"percentage":97},"Shell","#89e051",1,{"name":99,"color":100,"percentage":101},"Cython","#fedf5b",0,{"name":103,"color":104,"percentage":101},"C","#555555",{"name":106,"color":107,"percentage":101},"CMake","#DA3434",1110,228,"2026-04-02T03:14:28","NOASSERTION",4,"Linux, Windows, macOS","未说明（基于 PaddlePaddle，通常深度学习任务推荐 NVIDIA GPU，具体显存和 CUDA 版本需参考 PaddlePaddle 官方文档）","未说明（提及支持约 6600 氨基酸的超长单体蛋白预测，暗示需要较大内存）",{"notes":117,"python":118,"dependencies":119},"该工具基于百度飞桨（PaddlePaddle）深度学习平台。安装前置条件和详细指南请参阅项目中的 installation_guide.md 文件。部分功能（如 HelixFold3）最初发布时仅限非商业学术研究使用。支持超长单体蛋白（约 6600 个氨基酸）的结构预测。","3.6+",[75],[13],[122,123,124,125,126,127,128,129,130,131,132,133,134,135,136],"biocomputing","machine-learning","deeplearning","rna-structure-prediction","dti","representation-learning","graph-networks","protein-structure-prediction","self-supervised-learning","ppi","molecule-design","protein-folding","ddi","protein-docking","protein-design","2026-03-27T02:49:30.150509","2026-04-06T05:35:39.955989",[140,145,150,155,160,165],{"id":141,"question_zh":142,"answer_zh":143,"source_url":144},17907,"如何通过 pip 安装 PaddleHelix？遇到 'No matching distribution found' 错误怎么办？","如果直接运行 `pip install paddlehelix` 报错找不到版本，请尝试以下两种方法：\n1. 指定 python3 的 pip 模块运行：`python3 -m pip install paddlehelix`\n2. 检查并修改 anaconda 环境下的 pip 路径，确保使用的是正确 Python 环境对应的 pip。","https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Fissues\u002F221",{"id":146,"question_zh":147,"answer_zh":148,"source_url":149},17908,"运行 HelixFold-Single 时出现 'Operator fused_gate_attention does not have kernel' 或随机初始化参数导致的异常大数值错误，如何解决？","这通常是因为环境配置不匹配导致模型未能正确加载预训练权重（处于随机初始化状态）或缺少特定的算子内核。\n解决方案：\n1. 严格按照文档要求重新部署环境，确保 Python 版本为 3.7+。\n2. 必须使用官方指定的 PaddlePaddle 开发版（如 `paddlepaddle_gpu-0.0.0.post112`），官方正式版（如 2.3.2）可能不兼容。\n3. 推荐使用 NVIDIA 官方 Docker 镜像构建环境，以避免底层 CUDA\u002FcuDNN 依赖问题。","https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Fissues\u002F239",{"id":151,"question_zh":152,"answer_zh":153,"source_url":154},17909,"如何使用蛋白质预训练模型（Transformer\u002FLSTM\u002FResNet）获取蛋白质序列的特征表征（Embedding）？","可以通过源码中的 `tape_model.py` 实现。加载预训练模型后，利用 `hidden` 或 `pooled_hidden` 输出作为蛋白质的特征表征矩阵。\n具体逻辑参考：实例化模型类，输入蛋白质序列（如 \"MVLSPADKTNVKAAWGKVGAHAGEYG\"），提取中间层的隐藏状态即可用于后续的自定义操作，而不仅仅是获取分类结果。","https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Fissues\u002F64",{"id":156,"question_zh":157,"answer_zh":158,"source_url":159},17910,"预测长序列蛋白质（如长度超过 2000）时显存不足或运行失败，有什么优化建议？","对于长序列蛋白（例如 CASP14 中长度为 2180 的 T1044 蛋白已成功测试），建议采取以下措施：\n1. 开启 `unified_memory`（统一内存）选项。\n2. 开启 `DAP`（动态分配策略）。\n3. 如果条件允许，使用更大显存的 GPU（如 A100 80G）会更保险。","https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Fissues\u002F209",{"id":161,"question_zh":162,"answer_zh":163,"source_url":164},17911,"运行 Demo 时遇到报错，但日志不完整或使用了错误的 Paddle 版本，如何处理？","1. 如果遇到错误，请务必提供完整的错误日志（full error log）以便排查。\n2. 注意官方 PaddlePaddle 正式版（如 2.3.2）可能无法运行 HelixFold-Single。请卸载当前版本，并通过以下命令安装指定的开发版：\n`python -m pip install paddlepaddle-gpu==0.0.0.post112 -f https:\u002F\u002Fwww.paddlepaddle.org.cn\u002Fwhl\u002Flinux\u002Fgpu\u002Fdevelop.html`","https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Fissues\u002F216",{"id":166,"question_zh":167,"answer_zh":168,"source_url":169},17912,"在 Linux Docker 环境中运行代码遇到 ValueError 或 Kernel 缺失问题是什么原因？","这通常是由于 Docker 镜像选择不当或缺少特定的 GPU 内核支持。\n解决方案：建议使用 NVIDIA 官方提供的 Docker 镜像（nvidia docker），并确保在安装依赖时选择了正确的镜像版本。如果重建镜像后仍报错，请检查是否安装了与 CUDA 版本（如 11.2）严格匹配的 PaddlePaddle GPU 包。","https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleHelix\u002Fissues\u002F219",[171,175,180,185],{"id":172,"version":173,"summary_zh":78,"released_at":174},108201,"v1.2.2","2023-08-01T09:31:36",{"id":176,"version":177,"summary_zh":178,"released_at":179},108202,"v1.1.0","为分子表示学习算法 GEM 创建 DOI。","2021-12-15T01:46:45",{"id":181,"version":182,"summary_zh":183,"released_at":184},108203,"v1.0","1) 从静态框架更新为动态框架；2) 新增应用：分子生成和药物协同作用。","2021-07-09T06:04:32",{"id":186,"version":187,"summary_zh":188,"released_at":189},108204,"v1.0b","PaddleHelix 的第一个版本。PaddleHelix 是一个基于机器学习的生物计算框架，旨在推动疫苗设计、药物发现和精准医学等领域的发展。PaddleHelix 提供了化合物表示学习、蛋白质表示学习、药物-靶标相互作用预测以及 RNA 折叠等应用场景的示例。","2020-12-22T09:51:54"]