Skip to content

Limitations

Read this before relying on a result. Each item is a real constraint, not a disclaimer.

The model

  • The extracellular space is a fixed bath. There are no extracellular electric fields and no extracellular diffusion. This matches BETSE run without its is_ecm option.
  • There is no cell volume or osmotic regulation. A depolarized steady state is reached within seconds, but over hours its cells gain Na⁺, up to about 200 mM in the default cell, with a slowest time scale near 3 hours. Steady states of depolarized cells are therefore not physiological at long times.
  • The channels that produce bistability are phenomenological. They are CAM-type gating functions, not models of a named protein. The BETSE Kir2.1 model alone does not give bistability with the default profile.
  • There is no gene regulatory network. Voltage does not feed back onto expression.

The drug library

  • It contains 10 potency values from 4 primary sources. Each was measured in a heterologous expression system or on purified protein, in a named species. A tissue in an embryo is a different preparation.
  • Selectivity is not modelled. A drug acts on one target in the model. Known off-target actions are recorded in each entry and reported by selectivity_warnings, but not simulated. Digoxin is the clearest case: it blocks the Na/K-ATPase about a thousand times more potently than Cx43, so at a junction-blocking concentration the pump is blocked too.
  • Hill coefficients are mostly unpublished. Where none is given, 1 is used and the entry says so.
  • The Cx43 values come from a dye-transfer assay, which measures permeability to calcein rather than electrical coupling. These are different measurements.
  • Dihydropyridines are excluded on purpose. Their block is state-dependent, so one IC50 is wrong in a tissue whose cells sit at different voltages.
  • The screen in examples/drug_repurposing.py finds that no drug in the library normalizes a depolarized domain. Every compound with a verified value acts on a polarizing element, so blocking it depolarizes. That is a statement about which potency data are published, not about bioelectric medicine.

These are in-silico hypotheses, not medical advice.

Performance

  • Backpropagation through time costs more as the simulated interval grows. An 8×8 ion-level tissue takes about 4 s per optimization iteration for a 0.5 s rollout.
  • Steady-state design avoids that: a 16×16 tissue takes about 9 s per iteration regardless of how long the tissue would take to relax.

Scope of the gradients

steady_state_implicit differentiates the attractor the start state relaxes to. It does not include the jump that happens when a parameter change moves the start state into a different basin. Near a fold, read the dose response rather than the derivative.