PROTEUS-REPAIR evidence snapshot

Original experimental study: Tsuboyama K et al. Nature 620, 434–444 (2023). DOI: 10.1038/s41586-023-06328-6.

Candidate CSV: unchanged copy of gate9c_candidate_robustness.csv read from the project Google Drive.
WT_name and partner_mut identify the protein domain and partner substitution. tested counts distinct damaging substitutions; both counts those passing both correction methods. physical/isotonic count each method separately. partner_effect is the median standalone partner effect in kcal/mol, positive stabilizing. adjusted_p is the original conditional same-site Bonferroni-adjusted screening value, not a newly validated p-value.

This site does not report matched background base rates or interaction effects. The complete original event-label definitions have not been independently reconstructed.

The Python audit requires the existing orientation-level dataframe d in the original Colab. Run: %run -i gate9_audit.py
It writes only new timestamped outputs. Syntax checked; not yet executed on the complete source data. The block-column sensitivity test has its own exchangeability and between-block independence assumptions.