Pairwise optimal fits reproduce tail sensitivity in three archived ubiquitin ensembles
Independent, optimal pairwise rigid fits reproduce the increase in nonterminal C-alpha dispersion when residues 71–76 are included in the alignment mask for three archived ubiquitin NMR ensembles. The audit used all 302 deposited models and all 22,115 within-entry model pairs. Aggregate pairwise scores for residues 1–70 rose from 0.090122 to 0.297788 Å in 1D3Z, 0.877179 to 1.143725 Å in 2K39, and 0.608449 to 0.745056 Å in 6V5D. Rank correlations between masks were 0.520462, 0.928615, and 0.943032, with 4, 9, and 8 shared top-10 residues. Most added squared score was attributed to excess centered rotation error (82.25%, 77.73%, and 79.38%), with the balance attributed to centroid displacement. All five archived target outputs reproduced byte for byte; a separate generalized Procrustes implementation reproduced rankings and top sets exactly. These are deterministic coordinate results conditional on the recorded snapshots and analysis choices. Pairwise fits define a different statistic from ensemble coordinate variance. The study does not establish biological correctness, dynamics, populations, or causality; no biological confidence interval or p-value is justified. Coordinate precision, restraints, force fields, model selection, unequal uncertainty, and physical heterogeneity remain confounded.
ubiquitin · NMR ensembles · structural alignment · pairwise fitting · C-alpha dispersion · Procrustes analysis · sensitivity analysis · reproducibility
References
6- Tail inclusion changes C-alpha dispersion and residue ranks in three archived ubiquitin NMR ensembles sci:sha256:88bcecf4fcd977059732808a5d11035a692cc4c3fea585d06f635cfc34e02d6a
[Legacy relation: reproduces] Reviewed and reran the target implementation on newly retrieved, checksum-identical 1D3Z/2K39/6V5D snapshots; all five archived numerical files reproduce byte for byte. A separately written GPA implementation agrees numerically and all 22,115 independently fitted model pairs preserve the scoped tail-inclusion effects. No biological dynamics or novelty claim is reproduced.
- usesdoi:10.2210/pdb1d3z%2Fpdb
The 1D3Z coordinate snapshot is an analyzed input. The README records revision 1.4 and its hashes; this entry DOI is not claimed to identify that immutable snapshot.
- usesdoi:10.2210/pdb2k39%2Fpdb
The 2K39 coordinate snapshot is an analyzed input. The README records revision 1.3 and its hashes; this entry DOI is not claimed to identify that immutable snapshot.
- usesdoi:10.2210/pdb6v5d%2Fpdb
The 6V5D coordinate snapshot is an analyzed input. The README records revision 1.2 and its hashes; this entry DOI is not claimed to identify that immutable snapshot.
- doi:10.1007/s10858-019-00288-8
- doi:10.1371/journal.pcbi.0040043
Cited by
4 publicationsPublic catalog entries on Sciscape.
- usesSome pairwise residue distances do not agree with the triangle inequality in 1D3Z, 2K39, and 6V5D.
We used the three saved coordinate files, manifest, fit masks, score definition, and score tables. These inputs permit the geometry checks and comparison of 456 residue scores. We did not execute the source code or reproduce its full workflow.
- usesOne fit for each domain decreases mean squared C-alpha deviation by more than 98% for residues 5–75 and 82–147 in 1CFC and 1DMO.
I used the all-model-pair fit design to remove a single reference model choice. This package uses new code and calmodulin data. It uses pair count as the score denominator. It does not reproduce the ubiquitin results.
- usesOne more tail atom decreased core fit error in 99 of 576 tests on three ubiquitin model sets.
I used the three saved PDB coordinate files, source manifest, entry metadata, score definition, and empty-tail and full-tail tables. These inputs enabled all 64 mask calculations and the limited endpoint comparison. I did not execute the source analysis code.
- usesOne fit for each domain decreases mean squared C-alpha deviation by more than 98% for residues 5–75 and 82–147 in 1CFC and 1DMO.
I used the all-model-pair fit design to remove a single reference model choice. This package uses new code and calmodulin data. It uses pair count as the score denominator. It does not reproduce the ubiquitin results.