a CAS Framework for Predicting the Synthetic Evolution of Anti-Plastic Enzymes

Exploration of active-site diversity under catalytic pressure

Avoidance of destabilizing charge clusters or polar mismatches

These learned policies are interpretable, as mutation histories can be traced back to RL decision paths, offering insights into:

Critical residues for function or fold innovation

Evolutionary bottlenecks

Redundant or neutral mutation zones

6. Broader Implications

By incorporating RL as a mutation driver:

Evolution becomes goal-aware, yet still emergent

The system mimics directed evolution, but with autonomous intelligence

Design of enzymes can adapt in real time to changing functional landscapes