Problem
9-pathway experimental accuracy plateaus at ~70%. After connectivity is addressed, the remaining error is missing biological dynamics, not the propagation math.
Reactome doesn't encode:
- Emergent stress signals (DNA-damage level, ROS, hypoxia, …)
- Time-dependent decision logic (e.g. p53 / DSB-repair cascade timing)
~15–25 TP53 failures are structurally unfixable without virtual nodes representing these signals.
Direction
- Start with hand-coded heuristics for the most common missing signals.
- Graduate to ML-learned weights via the
parameter_learning.jl scaffolding.
See memory notes: missing-dynamics, sigmoid-design-intent.
Problem
9-pathway experimental accuracy plateaus at ~70%. After connectivity is addressed, the remaining error is missing biological dynamics, not the propagation math.
Reactome doesn't encode:
~15–25 TP53 failures are structurally unfixable without virtual nodes representing these signals.
Direction
parameter_learning.jlscaffolding.See memory notes: missing-dynamics, sigmoid-design-intent.