A machine learning framework can distinguish molecules made by biological processes from those formed through non-biological processes and could be used to analyze samples returned by current and future planetary missions. The findings are published in the journal PNAS Nexus.