New York University researchers have trained an AI model to learn chemical patterns associated with stability in drug-like molecules and accurately predict where their hydrogen atoms should be positioned.