[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-80755":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":15,"stars30d":13,"stars90d":15,"forks30d":15,"starsTrendScore":15,"compositeScore":14,"rankGlobal":10,"rankLanguage":10,"license":16,"archived":17,"fork":17,"defaultBranch":18,"hasWiki":17,"hasPages":17,"topics":19,"createdAt":10,"pushedAt":10,"updatedAt":20,"readmeContent":21,"aiSummary":22,"trendingCount":15,"starSnapshotCount":15,"syncStatus":23,"lastSyncTime":24,"discoverSource":25},80755,"EV-A71_2A_benchmark","OpenBind-Consortium\u002FEV-A71_2A_benchmark","OpenBind-Consortium","Benchmarking tools and analysis scripts for processing and evaluating structural data from the OpenBind EV-A71 2A dataset","https:\u002F\u002Fopenbind.uk\u002F",null,"Python",42,1,41,0,"Apache License 2.0",false,"main",[],"2026-06-12 04:01:29","# OpenBind EV-A71 2A dataset and benchmarks\n\nThis repository contains data and reference benchmarks for the first [OpenBind](https:\u002F\u002Fopenbind.uk\u002F) release: a structure–affinity dataset for structure-based AI.\n\nThe release includes experimentally determined protein–ligand complexes, affinity measurements, and benchmark evaluations for docking, cofolding, and affinity prediction methods.\n\nFor more background, see our accompanying blog post, [OpenBind’s first release: A structure–affinity dataset for structure-based AI](https:\u002F\u002Fopenbind.uk\u002Fnews\u002Fblog-openbinds-first-release-a-structure-affinity-dataset-for-structure-based-ai\u002F), which discusses the release, the EV-A71 2A protease target, and the benchmark results in more detail.\n\n## Overview\n\nThe dataset focuses on EV-A71 2A protease and contains:\n\n- 925 crystallographic binding events  \n- 699 compounds  \n- 601 compounds with affinity measurements  \n\nThis is a dense, single-target dataset designed for:\n- model training and fine-tuning  \n- benchmarking and comparison  \n- error analysis and method development  \n\nUnlike many public resources, it provides both structure and affinity across a coherent compound series, making it valuable for studying local structure–activity relationships.\n\n## What’s included\n\n- Curated structure and affinity data (see data links below)\n- Prepared benchmark tables for plotting and analysis\n- Reference evaluations across:\n  - docking (classical and ML-based)\n  - cofolding methods\n  - affinity prediction\n\n\n## Data and external resources\n\n- Blog post: [OpenBind’s first release: A structure–affinity dataset for structure-based AI](https:\u002F\u002Fopenbind.uk\u002Fnews\u002Fblog-openbinds-first-release-a-structure-affinity-dataset-for-structure-based-ai\u002F)\n- Dataset: [Zenodo](https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.20026661) \u002F [Fragalysis](https:\u002F\u002Ffragalysis.diamond.ac.uk\u002Fviewer\u002Freact\u002Fpreview\u002Ftarget\u002FA71EV2A\u002Ftas\u002Flb42888-1)  \n- Benchmarks: this repository  \n- Fine-tuned OpenFold3-p2 model: [of3p2-ft-ev2a.ckpt](https:\u002F\u002Fopenfold3-data.s3.amazonaws.com\u002Fopenfold3-parameters\u002Fopenbind\u002Fof3p2-ft-ev2a.ckpt). This model was fine-tuned only on the EV-A71 2A fragment-screen fine-tuning data and improves follow-on compound cofolding performance for this target. It likely has reduced performance on other targets, which we will address in future releases.\n- Experimental protocols: [OpenBind protocols.io workspace](https:\u002F\u002Fwww.protocols.io\u002Fworkspaces\u002Fopenbind) \n\n## Usage\n\nTo reproduce benchmark figures:\n\n```bash\npython plotting\u002Fplot_figures.py\n```\n\nSee [`plotting\u002FREADME.md`](https:\u002F\u002Fgithub.com\u002FOpenBind-Consortium\u002FEV-A71_2A_benchmark\u002Fblob\u002Fmain\u002Fplotting\u002FREADME.md) for details.\n\n\n## Citation and license\n\n- Repository license: [Apache 2.0](https:\u002F\u002Fgithub.com\u002FOpenBind-Consortium\u002FA71EV2A-benchmark\u002Fblob\u002Fmain\u002FLICENSE)\n- Data license: CC0 1.0 Universal  \n- DOI: [10.5281\u002Fzenodo.20026661](https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.20026661)  \n","该项目提供了一套用于处理和评估OpenBind EV-A71 2A数据集结构数据的基准工具和分析脚本。核心功能包括实验确定的蛋白-配体复合物、亲和力测量以及对接、共折叠和亲和力预测方法的基准评估。项目使用Python编写，支持模型训练与微调、基准测试与比较、错误分析及方法开发。适用于需要研究局部结构-活性关系的场景，特别是针对EV-A71 2A蛋白酶的研究。相比于许多公共资源，该项目提供了连贯化合物系列中的结构和亲和力数据，使其在结构基础的人工智能领域具有重要价值。",2,"2026-06-11 04:01:55","CREATED_QUERY"]