Pre-registration: inherited centriole age and interdivision timing in immortalized fibroblasts (ARGUS-OS1 protocol v0.3)

Pre-registration of the ARGUS-OS1 pilot protocol, deposited before any data exist: no images, no tracked lineages, no pilot run at the time of writing. It fixes, in a form a reader can check later, the hypotheses, the primary endpoint, the stopping criteria, the operational definition of the senescence endpoint, the sampling requirement, the sample size and the readout acceptance thresholds. Fixed here and not afterwards. Sampling at 100 nm/px or finer, no binning above 2x2 (basis: detection recall 0.78 at about 80 nm/px against 0.46 at about 250 nm/px). Analysis floor n = 115 cells per group and recruitment target 150, for a 2 h difference in interdivision interval at alpha = 0.05 two-sided and power 0.8 with a 6 h between-cell spread. Readout acceptance test on the frozen phantom before the first imaging run: localisation RMSE <= 0.6 px, recall >= 0.75, precision >= 0.90, and an interdivision-interval MAE <= 1.0 h that is not yet implemented; failing that test stops the pilot rather than warning about it. Stopping rules. The window closes on a significant OLD-versus-NEW divergence, or on the senescence endpoint defined by D1 and D2 with D3 confirmatory, or on the end of the window with the conclusion recorded as 'not detected in the window'. Arrest with D2 negative is reported as arrest of undetermined cause, not as senescence. What this is not. It carries no biological claim and no therapeutic claim. The calibration behind the sampling and power figures comes from a synthetic benchmark (DOI 10.5281/zenodo.24148459), which says nothing about whether inherited centriole age predicts division timing. Changes after this deposit are issued as dated amendments in a new version, never as edits.