Abstract
RNA is likely to be the first biomolecule to have appeared during evolution, but the abiotic synthesis of long oligonucleotides through phosphoester bond formation is an unsolved problem. Because the uncatalyzed reaction is extremely slow, experimental studies bring limited and indirect information on the molecular mechanism, the nature of which remains debated. We solve this issue by using neural network potentials systematically trained to explore, with enhanced sampling strategies, the chemical phase space for such complex reaction involving several proton transfers and exchanges of heavy atoms in explicit solvent, with quantum accuracy but moderate computational cost. A dissociative mechanism is thermodynamically favored over an associative one, with the formation of a metaphosphate transition state and direct participation of water solvent molecules. These observations rationalize unexplained experimental results and the temperature-dependence of the reaction rate, and they pave the way for the design of more efficient abiotic catalysts and activating groups.