Cross-validate 15-PGDH binder designs using Boltz-2 structure prediction.
npx skills add https://github.com/alex-hh/in-silico-pgdh --skill pgdh-boltz2
Please help me install this Agent Skill. Skill: pgdh_boltz2 Source: https://github.com/alex-hh/in-silico-pgdh/tree/main/.claude/skills/pgdh_boltz2 Command: npx skills add https://github.com/alex-hh/in-silico-pgdh --skill pgdh-boltz2
This Skill provides an independent cross-validation of protein binder designs for the 15-PGDH target, ensuring the reliability of computational predictions before experimental testing.
Use the pgdh_boltz2 skill to cross-validate binder designs for 15-PGDH using a provided YAML input file.
High-intent search queries and answers about installing and using this skill.
You cross-validate protein binder designs for 15-PGDH by running Boltz-2 structure prediction to generate confidence metrics like ipTM, pTM, and pLDDT. This serves as an orthogonal validation step to ensure prediction reliability before experimental testing.
The best way to validate RFdiffusion3 binders is using an independent model like Boltz-2 to predict complex structures. This orthogonal validation step generates confidence metrics such as ipTM and pLDDT to independently assess predicted binder-target interactions.
To run Boltz-2 structure prediction on the Lyceum platform, you need Lyceum authentication, an activated virtual environment, and uploaded scripts. You execute the prediction by providing a YAML input file containing your designed binder data.
Confidence metrics used to assess predicted protein binder interactions include ipTM, pTM, and pLDDT. These scores are generated by Boltz-2 structure prediction to evaluate the quality of designed binder-target complexes for 15-PGDH.
Yes, you need Lyceum authentication to generate pTM and ipTM scores for designed binders. Additionally, you must have an activated virtual environment and the necessary prediction scripts uploaded to the Lyceum platform for execution.
Independent structure prediction is used for protein design campaigns to ensure the reliability of computational predictions before experimental testing. By using a model independent from the initial design tools like BoltzGen, you can compare confidence scores to identify the most promising candidates.
alex-hhScore predicted protein-protein interactions using ipSAE, pDockQ, pDockQ2, and LIS metrics.
zongtingweiFilter protein-design candidates using structural, interface, and biophysical QC metrics.
junior1pGuide protein binder design workflows with BoltzGen, BindCraft, and Chai validation.
adaptyvbioPredict protein structures from FASTA sequences using Chai-1.
