ipsae

Rank binder designs using ipSAE scores from AlphaFold2, AlphaFold3, and Boltz1 predictions.

151|20|Updated Jan 15, 2026

Search Tags:#AlphaFold #ranking #protein-protein-interactions #binder-design #ipSAE #structure-prediction

Ready to run

npx skills add https://github.com/adaptyvbio/protein-design-skills --skill ipsae

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ipsae
Source: https://github.com/adaptyvbio/protein-design-skills/tree/main/skills/ipsae
Command: npx skills add https://github.com/adaptyvbio/protein-design-skills --skill ipsae

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

ipSAE-based ranking provides an objective, scalable method to prioritize protein-protein interaction designs for experimental validation by scoring predicted interfaces.

Core Features & Use Cases

Score designs across AlphaFold2, AlphaFold3, and Boltz predictions to rank binders by interface quality.
Compare predictors (ipSAE vs ipTM / iPAE) to select top candidates.
Batch ranking and reporting with chain-pair and residue-level scores for large design sets.
Use Case: Prioritize designs for experimental testing to improve hit rates and reduce bench time.

Quick Start

Run ipsae.py with your predictor outputs to generate scores for ranking designs, e.g., python ipsae.py scores_rank.json design_0.pdb 10 10

Frequently Asked Questions about ipsae

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I rank protein binder designs for experimental validation?▼

You can rank protein binder designs by calculating ipSAE scores from AlphaFold2, AlphaFold3, or Boltz1 predictions. This scoring method objectively prioritizes protein-protein interaction candidates to improve experimental hit rates and reduce bench time.

What is ipSAE scoring and how does it evaluate protein-protein interactions?▼

ipSAE scoring is a method that evaluates protein-protein interaction interfaces using predicted aligned error and structural data. It generates chain-pair and residue-level scores to objectively assess and compare the interface quality of predicted protein complexes.

Can I use ipSAE scores to compare AlphaFold2, AlphaFold3, and Boltz1 predictions?▼

Yes, you can compare ipSAE scores across AlphaFold2, AlphaFold3, and Boltz1 predictions. The skill processes PAE and structure inputs from these predictors to standardize ranking and help select top binder design candidates.

How do I run ipSAE scoring on my structure prediction outputs?▼

You run the ipsae.py Python script with your predictor outputs, such as running python ipsae.py scores_rank.json design_0.pdb 10 10. This generates chain-pair and residue-level scores for batch ranking large sets of structure prediction outputs.

Do I need Python and NumPy to calculate ipSAE scores for my designs?▼

Yes, you need Python and NumPy installed to calculate ipSAE scores. The script requires these dependencies to process PAE and structure input files and output chain-pair and residue-level scoring data.

How does ipSAE ranking compare to ipTM or iPAE for filtering binder designs?▼

ipSAE ranking provides an alternative to ipTM or iPAE by calculating interface quality scores specifically for comparing predictors and prioritizing binder designs. It scales objectively to filter outputs from structure prediction pipelines for experimental testing.