Skip to content

8. Gaps and ambiguities (questions for the author)

8. Gaps and ambiguities (questions for the author)

  • G1 — "every 100 fs" is impossible. T L280-281 says the 50 frames were taken from the final 5 ns "i.e. every 100 fs". 50 frames over 5 ns is every 100 ps, and coordinates were only written every 5 ps (T L273-274). Assume 100 ps. Evidence (§1.1): the archived PQR files are named …_aligned_100ps_NN, 99 of them, and the analysis notebook reads the DCD at dt=100. CLOSED [verified]1, §1.1: the DCD frames are the PQRs at 100 ps, in time order. The final 5 ns are PQRs 50–99 (DCD frames 48–97).
  • G2 — 25 % isolevel is unjustified. No rationale, no sensitivity study, no comparison against the solvent-excluded surface. Since pore radius enters conductance roughly as r^2, this is the single largest untested lever in the geometry. Ask what other isolevels were tried. Measured on 2WCD, 28 September 2026 (§1.4): ±0.1 of isolevel moves the bulk-resistor conductance by about ±5 %.
  • G3 — the manual vertex edit is unreproducible. The final boundary polyline is not published and its vertex count is not stated; the 30-degree-wedge claim is qualitative. Without the vertex list a bit-for-bit geometry match is impossible; treat +-1 % on conductance as the tolerance floor. For VAL-05 the list is delivered, and its construction (pqr2grid binning at L = 15 nm, then a hand edit that moved vertices; §1.2, §1.4) puts ±1 % out of reach. The author's tolerances are in the SPECIFICATION.md §7.4 NOTE on VAL-05.
  • G4 — q_i/(2 pi r_i) vs /(2 pi r). The prose (T L404-408) divides by the atom's radial position r_i; the implementation (COMSOL, knowledge/01-physics-epnpns/05-fixed-charge-from-structure.md) divides by the field point's r. These differ within the Gaussian's support and diverge differently near the axis. The COMSOL form is what produced -72.9 e, so implement that; but the discrepancy should be confirmed.
  • G5 — CLOSED [verified]2, §3.1. The file stores the sum without e, in e/m^2: its planar integral is the integer -72, to 4.7e-12. e_const in the COMSOL assembly supplies the coulombs exactly once, and the e printed inside eq. eq:scdpore is not also in the table.
  • G6 — pore averages omit the 2 pi r Jacobian. eq:pore_surface_integral integrates beta * X dr dz, an unweighted area average in the (r,z) half-plane, while the main text calls it an average "over the entire pore volume" (T L823). A true volume average needs 2 pi r dr dz. Every number in §6.3 that is a pore average (Na+/Cl- ratios, PS/PB charges) depends on this choice; the wall-adjacent PS region is systematically under-weighted without the Jacobian. Must be resolved before those become regression targets. Related: <Q_ion>_PS and <Q_ion>_PB are quoted in units of e (i.e. counts, hence genuine volume integrals), not averages, despite the < > notation.
  • G7 — radial potential at 0.15 M is internally inconsistent. T L1117-1118 states the lumen value as "approx -14 mV" (and that it is below 1 kT/e) and the constriction as "approx -47 mV", but tab:radial_potential gives -29 mV and -57 mV at 0.15 M. -29 mV is above 1 kT/e, so the prose claim fails against its own table. The -14/-30 pair matches the 0.5 M table row, so the text may have been written against a different row. Trust the table.
  • G8 — constriction z-range sign typo. T L1043 prints "1.85 < z < 1.6 nm"; everywhere else it is -1.85 < z < 1.6 nm (T L881, L1076, TA). Use -1.85.
  • G9 — bilayer placement in z is never stated. Thickness 2.8 nm is given but its z-centre is not. The pore spans -1.85 to 12.25 nm and the constriction is symmetric about z ~ 0, so a bilayer centred at z = 0 (spanning -1.4 to +1.4 nm) is the obvious reading, but it is an inference. Compounding this, the MD/electrostatics frame places the structure's centre of mass at (0, 0, 55 Angstrom) (TA eq:internal_radius discussion) — the shift from MD coordinates to model coordinates is not given anywhere. Evidence (§1.1): the protein-only MD archive carries no lipids, so the bilayer cannot fix it. But the MD frame's all-atom extent brackets the model's pore extent to within 0.3–0.4 nm at each end, which is consistent with an offset near zero. CLOSED by the author: the MD trajectory was centred on the middle of the bilayer, so ClyA's centre_z_nm is 0 in the MD frame.
  • G10 — CLOSED [verified]2, §3.1. r in [0, 7] nm, z in [-3.5, 13.5] nm at 0.005 nm: 1401 x 3401 = 4.76e6 nodes, the order the estimate predicted.
  • G11 — nothing on mesh or solver tolerances. See §4 and §5. No convergence study of any kind is reported, so the published numbers carry no stated discretisation error bar.
  • G12 — MD ion parameters unspecified beyond "CHARMM36" (no NBFIX / Beglov-Roux statement). Irrelevant to the continuum solve, needed only if the MD is rerun.
  • G13 — "average between cis and trans" arithmetic. T L770-774 says the measured 0.66 corresponds to the average of 0.57 (1 M) and 0.84 (0.15 M); the arithmetic mean is 0.705, and for P_Na the mean of 1.3 and 5.4 is 3.35 versus the measured 1.9. The claim is qualitative, not arithmetic.

  1. git show 0ff2d619:docs/plans/wp18-structure-ingestion.md, Design §4 (MD archive frame spacing and PQR correspondence). ↩

  2. git show 0ff2d619:docs/plans/wp28-charge-deposition.md, Design §1 (prod5_clya_charge grid dimensions and integral). ↩↩