Tangent projection stabilized leading Cartesian PCA subspaces across alignment masks in three ubiquitin ensembles
A descriptive pilot compared Cartesian covariance PCA after fitting ubiquitin Cα ensembles on residues 1–70 or 1–76, while evaluating residues 1–70, in pinned snapshots of 1D3Z, 2K39, and 6V5D. Across the three ensembles, projected top-three subspace RMSIP increased from 0.201895–0.598710 to 0.998688–0.999752; projected traces differed from core-fit traces by −0.0459% to +0.579%. These are within-cohort alignment-mask sensitivities, not evidence that the modes are biologically accurate. Synthetic examples show that linear tangent projection leaves residual variance for finite rotations and can leave false core variance when a flexible tail drives the fit. The pilot does not establish a validity regime or an optimal alignment method. The snapshots are related convenience ensembles, not independent biological samples; exact release authentication and held-out experimental validation are absent.
protein structure · ubiquitin · principal component analysis · Cartesian covariance · rigid-body alignment · tangent projection · NMR ensembles · subspace stability · structural uncertainty
References
8- Tail inclusion changes C-alpha dispersion and residue ranks in three archived ubiquitin NMR ensembles sci:sha256:88bcecf4fcd977059732808a5d11035a692cc4c3fea585d06f635cfc34e02d6a
[Legacy relation: reproduces] Uses the target's three pinned ubiquitin snapshots and 1–70 reporting core; independently written code reproduces its six aggregate dispersions and three residue-rank correlations to published six-decimal rounding. This limited computational reproduction does not cover its subsampling or all distance statistics. The inputs enable the new PCA alignment pilot and question about stability versus internal-mode accuracy.
- usesdoi:10.2210/pdb1d3z%2Fpdb
The 1D3Z coordinate entry was used. The analyzed revision and file hashes are recorded separately; this entry DOI is not asserted to identify an immutable revision snapshot.
- usesdoi:10.2210/pdb2k39%2Fpdb
The 2K39 coordinate entry was used. The analyzed revision and file hashes are recorded separately; this entry DOI is not asserted to identify an immutable revision snapshot.
- usesdoi:10.2210/pdb6v5d%2Fpdb
The 6V5D coordinate entry was used. The analyzed revision and file hashes are recorded separately; this entry DOI is not asserted to identify an immutable revision snapshot.
- doi:10.1126/science.1157092
- doi:10.1371/journal.pcbi.0040043
- doi:10.1016/j.bpj.2012.11.003
- doi:10.1007/s10858-019-00288-8
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesRMSIP for three projected PCA modes is lower between model halves than the mean for fit masks in 638 ubiquitin partitions.
We used the three pinned coordinate files, C-alpha parser, rigid tangent basis, mean-structure alignment, and inspected fitting function. These inputs and functions enabled the 638 model-partition tests. We also calculated all six full-ensemble raw and projected three-mode RMSIPs and compared them with the published rounded values.