New code gives 6V5D minimum error counts of (3, 2) after the specified 0.00025 Å axis changes.

An independent implementation reproduces the published minimum contact error counts for PDB ubiquitin ensembles 1D3Z, 2K39, and 6V5D, plus all 42 specified sensitivity rows. For 6V5D, the baseline counts are (2, 3) for alignments on residues 1–70 and 1–76; a specified construction that changes four C-alpha residue pairs by 0.00025 Å on each coordinate axis gives (3, 2). The construction changes no individual 8 Å model contacts or majority labels, and the changed coordinates round to the deposited three-decimal values. The run gives 96 checks with zero failures, while preserving an earlier failed verifier and its diagnosis. This is a numerical sensitivity result, not a new biological experiment, a physical uncertainty model, or evidence about protein motion. The coordinate changes have no bond geometry constraints; the best threshold is selected against the same majority labels used for scoring; and the tested initial references do not establish a global alignment optimum.

ubiquitin · protein structure · NMR ensembles · C-alpha contacts · coordinate means · structural alignment · numerical sensitivity · reproducibility