Each round starts from the current champion enzyme. Mutate spawns 3–5 variants whose active-site geometry (shape fit) and electrostatics (charge fit) are nudged by a random amount — some mutations help binding, most are neutral or mildly harmful, a few are ruinous. Every variant's rate is computed from real Michaelis–Menten kinetics, not just decoration:
v = Vmax·[S] / (Km + [S])
Vmax = kcat·[E]total
kcat ∝ shape fit (how well the pocket cradles the substrate)
Km ∝ 1 / charge fit (how tightly it's held once bound)
Select best screens the library and promotes the variant with the highest reaction velocity at the current substrate concentration — exactly the "mutate → screen → select" loop real directed-evolution labs run for dozens of rounds (the method Frances Arnold won the 2018 Nobel Prize in Chemistry for). The chart tracks the champion's activity round by round: it should climb in bursts as beneficial mutations get fixed, then plateau as the pocket approaches its geometric optimum.
- Shape fit — how closely the active-site pocket's contour matches the substrate's shape (lock-and-key / induced fit); raises kcat.
- Charge fit — how well the pocket's electrostatic residues complement the substrate's polar groups; lowers Km (tighter binding).
- [S] substrate concentration — slide it up toward saturation to see v approach Vmax; slide it down toward Km to see how much binding affinity (not just raw speed) matters.