How it works

Protein language models trained on structure are probed for electrostatic signals. No extra training is needed; the knowledge is already encoded.