Traj Desk - notice

The app's agent prompt is derived from the agent skill "molecular-dynamics"
(@k-dense-ai/molecular-dynamics, skill version 1.1, author Kuan-lin Huang), section "Trajectory
Analysis with MDAnalysis" (load_trajectory, compute_rmsd, compute_rmsf, analyze_contacts) and its
best practices, in the repository K-Dense-AI/scientific-agent-skills by K-Dense Inc.
  https://github.com/K-Dense-AI/scientific-agent-skills (skills/molecular-dynamics)
The skill's front matter declares the MIT license. No text of the skill is redistributed verbatim;
the prompt was rewritten for this app.

The in-browser analysis (pdbtraj.js, mdcore.js, trajkit.js) is an independent JavaScript
implementation of what MDAnalysis 2.9.0 does for these calls: the PDB topology parser and PDB
coordinate reader (first model as topology, one frame per MODEL, float32 coordinates, masses from
the element column or guessed from atom names), the selection keywords listed on the page (with
MDAnalysis' equal precedence for and/or), align.AlignTraj(u, u, select=..., in_memory=True),
rms.RMSD, rms.RMSF (Welford), AtomGroup.radius_of_gyration() and distance-based contacts. The
optimal rotation is found with the quaternion eigenvector method rather than MDAnalysis' QCP code;
both solve the same least-squares problem. The mass table, the element-guessing tables and the
protein and nucleic residue-name lists are copied from MDAnalysis 2.9.0 (guesser/tables.py and
core/selection.py; MDAnalysis 2.9.0 is packaged under the GNU LGPL v3 or later, and its tables are
reproduced here as data). No other MDAnalysis, NumPy or SciPy code is included.

It was checked against MDAnalysis 2.9.0 running the skill's recipes on 700 random trajectories
(1-30 frames; mixed residues including HSD, MSE and CYX; ligands, water and ions; altlocs; segids
or chain ids; CRYST1 records; with and without element columns; 19 kinds of selection; random
reference frames, start frames, frame spacings and cutoffs) and on the 38 NMR models of PDB 1L2Y
with 21 settings - about 150,000 compared values. RMSD, RMSF, per-residue RMSF and radius of
gyration agree to within 2e-5 Angstrom and contacts exactly, except for selections of two atoms,
whose orientation is undefined (MDAnalysis then returns an arbitrary rotation); the page refuses
selections of fewer than three atoms or of atoms on one line. Two behaviours reproduced on purpose:
AlignTraj has no reference-frame argument (it aligns to frame 0), and near-zero RMSD values carry
MDAnalysis' QCP iteration noise of up to about 1e-3 Angstrom, which the reproduction check avoids.

Defects in the skill's snippets that this app reports and works around:
  compute_rmsf indexes R.results.rmsf (one value per selected atom) with universe atom indices
  (res_atoms.indices): for the default "backbone" selection of an all-atom protein it raises
  IndexError (confirmed with MDAnalysis 2.9.0 on PDB 1L2Y: "index 92 is out of bounds for axis 0
  with size 80").
  analyze_contacts calls contacts.contact_matrix(protein.positions, ligand.positions, radius), but
  MDAnalysis 2.9.0 defines contact_matrix(d, radius, out=None) on a precomputed distance matrix, so
  the call raises TypeError.

Example data (wwPDB, CC0 1.0 - https://www.wwpdb.org/about/usage-policies):
  trpcage-1l2y.txt: the 38 models of PDB entry 1L2Y, coordinates unmodified (header records
  trimmed). Neidigh, J.W., Fesinmeyer, R.M. & Andersen, N.H. (2002). Designing a 20-residue
  protein. Nat. Struct. Biol. 9, 425. https://doi.org/10.1038/nsb798
  trpcage-wrapped.txt: the same models with residues 11-20 of models 26-38 shifted by +40 A in x
  and a 40 A CRYST1 record per model, to imitate a periodic wrap. Altered on purpose; not data.
  streptavidin-biotin-synthetic.txt: a SYNTHETIC 30-frame trajectory built from PDB entry 1STP
  (chain A and the biotin, water removed) by adding random rigid-body motion, residue and atom
  displacements, a moving 45-52 loop, a slow hinge motion of residues 65-83 and the biotin moving
  out of the pocket from frame 18. It is not a simulation. Weber, P.C., Ohlendorf, D.H.,
  Wendoloski, J.J. & Salemme, F.R. (1989). Structural origins of high-affinity biotin binding to
  streptavidin. Science 243, 85-88 (PMID 2911722).

MDAnalysis:
  Michaud-Agrawal, N., Denning, E.J., Woolf, T.B. & Beckstein, O. (2011). MDAnalysis: A Toolkit
  for the Analysis of Molecular Dynamics Simulations. J. Comput. Chem. 32, 2319-2327.
  https://doi.org/10.1002/jcc.21787
  Gowers, R.J. et al. (2016). MDAnalysis: A Python package for the rapid analysis of molecular
  dynamics simulations. Proceedings of the 15th Python in Science Conference, 98-105.
  https://doi.org/10.25080/majora-629e541a-00e
  QCP (used by MDAnalysis): Theobald, D.L. (2005). Acta Cryst. A 61, 478-480; Liu, P.,
  Agrafiotis, D.K. & Theobald, D.L. (2010). J. Comput. Chem. 31, 1561-1563.