A recursive enzymatic competition network capable of multitask molecular information processing

Living cells understand their environment by combining, integrating and interpreting chemical and physical stimuli. Despite considerable advances in the design of enzymatic reaction networks that mimic hallmarks of living systems, these approaches lack the complexity to fully capture biological information processing. Here we introduce a scalable approach to design complex enzymatic reaction networks capable of reservoir computation based on recursive competition of substrates. This protease-based network can perform a broad range of classification tasks based on peptide and physicochemical inputs and can simultaneously perform an extensive set of discrete and continuous information processing tasks. The enzymatic reservoir can act as a temperature sensor from 25 °C to 55 °C with 1.3 °C accuracy, and performs decision-making, activation and tuning tasks common to neurological systems. We show a possible route to temporal information processing and a direct interface with optical systems by demonstrating the extension of the network to incorporate sensitivity to light pulses. Our results show a class of competition-based molecular systems capable of increasingly powerful information-processing tasks.