[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72207":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":16,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":10,"rankLanguage":10,"license":22,"archived":23,"fork":23,"defaultBranch":24,"hasWiki":25,"hasPages":23,"topics":26,"createdAt":10,"pushedAt":10,"updatedAt":30,"readmeContent":31,"aiSummary":32,"trendingCount":16,"starSnapshotCount":16,"syncStatus":33,"lastSyncTime":34,"discoverSource":35},72207,"boltz","jwohlwend\u002Fboltz","jwohlwend","Official repository for the Boltz biomolecular interaction models","",null,"Python",4027,840,65,114,0,10,21,68,30,30.77,"MIT License",false,"main",true,[27,28,29],"binding-prediction","drug-discovery","protein-structure","2026-06-12 02:03:00","\u003Cdiv align=\"center\">\n  \u003Cdiv> \u003C\u002Fdiv>\n  \u003Cimg src=\"docs\u002Fboltz2_title.png\" width=\"300\"\u002F>\n  \u003Cimg src=\"https:\u002F\u002Fmodel-gateway.boltz.bio\u002Fa.png?x-pxid=bce1627f-f326-4bff-8a97-45c6c3bc929d\" \u002F>\n\n[Boltz-1](https:\u002F\u002Fdoi.org\u002F10.1101\u002F2024.11.19.624167) | [Boltz-2](https:\u002F\u002Fdoi.org\u002F10.1101\u002F2025.06.14.659707) |\n[Slack](https:\u002F\u002Fboltz.bio\u002Fjoin-slack) \u003Cbr> \u003Cbr>\n\u003C\u002Fdiv>\n\n\n\n![](docs\u002Fboltz1_pred_figure.png)\n\n\n## Introduction\n\nBoltz is a family of models for biomolecular interaction prediction. Boltz-1 was the first fully open source model to approach AlphaFold3 accuracy. Our latest work Boltz-2 is a new biomolecular foundation model that goes beyond AlphaFold3 and Boltz-1 by jointly modeling complex structures and binding affinities, a critical component towards accurate molecular design. Boltz-2 is the first deep learning model to approach the accuracy of physics-based free-energy perturbation (FEP) methods, while running 1000x faster — making accurate in silico screening practical for early-stage drug discovery.\n\nAll the code and weights are provided under MIT license, making them freely available for both academic and commercial uses. For more information about the model, see the [Boltz-1](https:\u002F\u002Fdoi.org\u002F10.1101\u002F2024.11.19.624167) and [Boltz-2](https:\u002F\u002Fdoi.org\u002F10.1101\u002F2025.06.14.659707) technical reports. To discuss updates, tools and applications join our [Slack channel](https:\u002F\u002Fboltz.bio\u002Fjoin-slack).\n\n## Installation\n\n> Note: we recommend installing boltz in a fresh python environment\n\nInstall boltz with PyPI (recommended):\n\n```\npip install boltz[cuda] -U\n```\n\nor directly from GitHub for daily updates:\n\n```\ngit clone https:\u002F\u002Fgithub.com\u002Fjwohlwend\u002Fboltz.git\ncd boltz; pip install -e .[cuda]\n```\n\nIf you are installing on CPU-only or non-CUDA GPus hardware, remove `[cuda]` from the above commands. Note that the CPU version is significantly slower than the GPU version.\n\n## Inference\n\nYou can run inference using Boltz with:\n\n```\nboltz predict input_path --use_msa_server\n```\n\n`input_path` should point to a YAML file, or a directory of YAML files for batched processing, describing the biomolecules you want to model and the properties you want to predict (e.g. affinity). To see all available options: `boltz predict --help` and for more information on these input formats, see our [prediction instructions](docs\u002Fprediction.md). By default, the `boltz` command will run the latest version of the model.\n\n\n### Binding Affinity Prediction\nThere are two main predictions in the affinity output: `affinity_pred_value` and `affinity_probability_binary`. They are trained on largely different datasets, with different supervisions, and should be used in different contexts. The `affinity_probability_binary` field should be used to detect binders from decoys, for example in a hit-discovery stage. Its value ranges from 0 to 1 and represents the predicted probability that the ligand is a binder. The `affinity_pred_value` aims to measure the specific affinity of different binders and how this changes with small modifications of the molecule. This should be used in ligand optimization stages such as hit-to-lead and lead-optimization. It reports a binding affinity value as `log10(IC50)`, derived from an `IC50` measured in `μM`. More details on how to run affinity predictions and parse the output can be found in our [prediction instructions](docs\u002Fprediction.md).\n\n## Authentication to MSA Server\n\nWhen using the `--use_msa_server` option with a server that requires authentication, you can provide credentials in one of two ways. More information is available in our [prediction instructions](docs\u002Fprediction.md).\n \n## Evaluation\n\n⚠️ **Coming soon: updated evaluation code for Boltz-2!**\n\nTo encourage reproducibility and facilitate comparison with other models, on top of the existing Boltz-1 evaluation pipeline, we will soon provide the evaluation scripts and structural predictions for Boltz-2, Boltz-1, Chai-1 and AlphaFold3 on our test benchmark dataset, and our affinity predictions on the FEP+ benchmark, CASP16 and our MF-PCBA test set.\n\n![Affinity test sets evaluations](docs\u002Fpearson_plot.png)\n![Test set evaluations](docs\u002Fplot_test_boltz2.png)\n\n\n## Training\n\n⚠️ **Coming soon: updated training code for Boltz-2!**\n\nIf you're interested in retraining the model, currently for Boltz-1 but soon for Boltz-2, see our [training instructions](docs\u002Ftraining.md).\n\n\n## Contributing\n\nWe welcome external contributions and are eager to engage with the community. Connect with us on our [Slack channel](https:\u002F\u002Fboltz.bio\u002Fjoin-slack) to discuss advancements, share insights, and foster collaboration around Boltz-2.\n\nOn recent NVIDIA GPUs, Boltz leverages the acceleration provided by [NVIDIA  cuEquivariance](https:\u002F\u002Fdeveloper.nvidia.com\u002Fcuequivariance) kernels. Boltz also runs on Tenstorrent hardware thanks to a [fork](https:\u002F\u002Fgithub.com\u002Fmoritztng\u002Ftt-boltz) by Moritz Thüning.\n\n## License\n\nOur model and code are released under MIT License, and can be freely used for both academic and commercial purposes.\n\n\n## Cite\n\nIf you use this code or the models in your research, please cite the following papers:\n\n```bibtex\n@article{passaro2025boltz2,\n  author = {Passaro, Saro and Corso, Gabriele and Wohlwend, Jeremy and Reveiz, Mateo and Thaler, Stephan and Somnath, Vignesh Ram and Getz, Noah and Portnoi, Tally and Roy, Julien and Stark, Hannes and Kwabi-Addo, David and Beaini, Dominique and Jaakkola, Tommi and Barzilay, Regina},\n  title = {Boltz-2: Towards Accurate and Efficient Binding Affinity Prediction},\n  year = {2025},\n  doi = {10.1101\u002F2025.06.14.659707},\n  journal = {bioRxiv}\n}\n\n@article{wohlwend2024boltz1,\n  author = {Wohlwend, Jeremy and Corso, Gabriele and Passaro, Saro and Getz, Noah and Reveiz, Mateo and Leidal, Ken and Swiderski, Wojtek and Atkinson, Liam and Portnoi, Tally and Chinn, Itamar and Silterra, Jacob and Jaakkola, Tommi and Barzilay, Regina},\n  title = {Boltz-1: Democratizing Biomolecular Interaction Modeling},\n  year = {2024},\n  doi = {10.1101\u002F2024.11.19.624167},\n  journal = {bioRxiv}\n}\n```\n\nIn addition if you use the automatic MSA generation, please cite:\n\n```bibtex\n@article{mirdita2022colabfold,\n  title={ColabFold: making protein folding accessible to all},\n  author={Mirdita, Milot and Sch{\\\"u}tze, Konstantin and Moriwaki, Yoshitaka and Heo, Lim and Ovchinnikov, Sergey and Steinegger, Martin},\n  journal={Nature methods},\n  year={2022},\n}\n```\n","Boltz 是一个用于生物分子相互作用预测的模型家族。其核心功能包括通过深度学习技术预测蛋白质结构和结合亲和力，其中 Boltz-2 模型在联合建模复杂结构和结合亲和力方面超越了 AlphaFold3 和 Boltz-1，接近基于物理的自由能扰动（FEP）方法的准确性，同时运行速度提高了 1000 倍。这使得在早期药物发现中进行准确的计算机筛选成为可能。项目使用 Python 编写，并提供 CUDA 支持以加速计算。适用于需要高精度生物分子相互作用预测的场景，如药物发现和蛋白质工程。所有代码和权重均在 MIT 许可下免费提供，支持学术和商业用途。",2,"2026-06-11 03:40:51","high_star"]