The 2K39 and 6V5D majority contact graphs remain fixed at 8 Å within ±0.0005 Å coordinate limits
A partial numerical reproduction of two ubiquitin contact-graph summaries tested whether the reported eight differences could result from coordinate display precision. Using all 116 models in RCSB entry 2K39 and all 176 models in 6V5D, we reproduced contact counts and graph summaries at 20 settings, and primary pair counts and distance summaries. At the primary setting (chain A, author residues 1–70, sequence separation greater than 2, and an 8 Å cutoff), the majority graphs have 167 and 169 contacts, share 164, and differ by eight pairs. All 2,278 pair labels per graph remain fixed under independent closed coordinate cubes of ±0.0005 Å per axis; all eight differences remain. Individual model contact labels can change in 36 of 665,176 tested model-pairs. Across the 20-setting grid, whole graphs remain fixed at 10 settings; 18 rows at the other settings contain five distinct unresolved entry, cutoff, and pair combinations. These bounds use a display-precision assumption, not a physical-error estimate or a claim about the archive rounding rule. The result excludes the stated display precision as a cause of the primary graph differences. It does not identify their physical cause, select a better ensemble, or establish contact frequencies as biological probabilities. The structures share experimental information, and deposited models are not independent organism samples. The reproduction excludes other source structures, alignments, figures, and experimental fits.
ubiquitin · protein structure · contact maps · contact graphs · coordinate precision · structural ensembles · majority contacts · RCSB PDB · numerical reproduction
References
4- Deposited-model and cutoff sensitivity of ubiquitin Cα contact maps sci:sha256:48903fcc23599e9ca1df58d55fee49705b5ce8c4dd81ebe1cf50a5d6abb4ea7d
[Legacy relation: reproduces] The new code reproduces the two-entry contact and graph summaries at all 20 settings. It also reproduces the primary pair counts and distance summaries for 2K39 and 6V5D. The three reference tables supply the numeric comparison. The reproduction excludes the other entries, alignment results, figures, and experimental fits.
- doi:10.1126/science.1157092
- doi:10.1007/s10858-019-00288-8
- doi:10.1002/prot.22187
Cited by
5 publicationsPublic catalog entries on Sciscape.
- usesThe method changes the 6V5D minimum error counts from (2, 3) to (3, 2) with 0.00025 Å axis limits.
The method uses exact integer distance limits for coordinate boxes to keep the majority labels fixed in the new alignment tests. This run also gives these limits for 1D3Z.
- A minimum of 50 model removals makes the tested 2K39 and 6V5D contact graphs agree.
[Legacy relation: reproduces] The executed coordinate-rounding check gave the same 167 and 169 contacts, eight graph differences, and fixed majority labels for all 2,278 pairs per entry. This relation covers only the two-entry primary 8 Å result. It does not cover the full distance-limit grid.
- At 8 Å for residues 1–70, 2K39 and 6V5D contact graphs can agree with model weight ratios ≤3 but not ≤2.
[Legacy relation: reproduces] New code gives the same 4,556 primary C-alpha contact counts and 9,112 coordinate-bound counts for 2K39 and 6V5D at 8 Å. The prior reference table supplies this comparison. This check includes chain A, residues 1–70, sequence separation greater than 2, and contact mass at least one half. It does not cover the full cutoff grid, other entries, or alignment results.
- usesNew code gives 6V5D minimum error counts of (3, 2) after the specified 0.00025 Å axis changes.
The code uses the integer coordinate-box distance-bound method. It shows that all 2,278 majority labels per entry stay the same under independent ±0.0005 Å axis limits. This permits comparison of coordinate means against a fixed target graph.
- usesIn one 6V5D test, bond rotations within 0.0005 Å axis limits reduce the 1–76 minimum contact errors from three to two
The integer coordinate-box distance limits show that all 2,278 majority labels stay fixed within independent ±0.0005 Å axis limits for 6V5D. This permits the comparison against fixed majority labels.