Stage 03

Kinetic target engagement (cellTE)

The third stage does not ask about affinity in isolation. It asks what fraction of the target actually ends up occupied inside a living cell.

Distribution of cellTE scores and the top candidates after rescoring
Target engagement in the cell: binding kinetics and rebinding.

A two-pool model#

The free cytoplasmic pool and the target-bound pool are described by differential equations balancing three processes: entry through the membrane, efflux, and association with and dissociation from the target (KD = koff / kon). The equations are integrated with the LSODA solver.

Inputs from stage 2#

Entry and accumulation indicators from stage 2 are converted into an inbound permeability and an efflux rate by heuristic, uncalibrated functions. In the first pass, binding kinetics is taken as a fixed prior, because affinity is not yet known.

Two passes and one gate#

The engagement read from the simulation sets the order of candidates entering structure prediction. After that, the model runs again with the predicted dissociation constant in place of the prior. Only this second pass acts as a gate: a minimum engagement threshold and a maximum KD of 100 µM.

Agreement between the two passes is not an independent validation, because both share the same permeability and efflux mappings.