The checked 6V5D records do not give the original model-to-group map.

This contribution asks what fraction of C-alpha contact-distance variance in 6V5D lies between the original 22 refinement groups of eight models. The checked PDB header, mmCIF fields, RCSB entry record, paper, and supplement did not provide the original model-to-group map; the search does not establish that no such map exists. Using 2026-10-09 snapshots of 6V5D revision 1.2 (176 models) and 2K39 revision 1.3 (116 models), the analysis reproduces a set of 69 shared invariant C-alpha contacts among 2,278 eligible pairs. It tests five illustrative 22-by-8 maps of 6V5D. The between-group fraction for the 69 contacts ranges from 3.255710% to 38.586309%; these constructed maps do not identify the original fraction, which remains unknown and could lie outside that range. For 1,000 uniform random equal-size partitions, the mean fraction is 12.018564% (Monte Carlo standard error 0.021400 percentage points), near the exact fixed-data expectation of 12%. This reference is not a biological null model. Checks support the arithmetic and parser outputs, and a separate run by the same researcher reproduced the output files; this is not independent verification. The selected, dependent models and coordinates do not establish physical state probabilities, biological causes, or the variance split for the original refinement groups.

ubiquitin · protein structure · NMR ensemble · PDB · C-alpha distance · variance decomposition · refinement groups · reproducibility