22Synthetic biological circuits are promising tools for developing sophisticated systems for medical, 23 industrial, and environmental applications. So far, circuit implementations commonly rely on gene 24 expression regulation for information processing using digital logic. Here, we present a new 25 approach for biological computation through metabolic circuits designed by computer-aided tools, 26implemented in both whole-cell and cell-free systems. We first combine metabolic transducers to 27 build an analog adder, a device that sums up the concentrations of multiple input metabolites. 28Next, we build a weighted adder where the contributions of the different metabolites to the sum 29 can be adjusted. Using a computational model trained on experimental data, we finally implement 30 two four-input "perceptrons" for desired binary classification of metabolite combinations by 31 applying model-predicted weights to the metabolic perceptron. The perceptron-mediated neural 32 computing introduced here lays the groundwork for more advanced metabolic circuits for rapid 33 and scalable multiplex sensing. 34Introduction 2 Living organisms are information-processing systems that integrate multiple input signals, 3 perform computations on them, and trigger relevant outputs. The multidisciplinary field of 4 synthetic biology has combined their information-processing capabilities with modular 5 and standardized engineering approaches to design sophisticated sense-and-respond 6 behaviors 1-3 . Due to similarities in information flow in living systems and electronic 7 devices 4 , circuit design for these behaviors has often been inspired by electronic circuitry, 8 with substantial efforts invested in implementing logic circuits in living cells 4-6 . 9 Furthermore, synthetic biological circuits have been used for a range of applications 10 including biosensors for detection of pollutants 7,8 and medically-relevant biomarkers 9,10 , 11 smart therapeutics 11,12 , and dynamic regulation and screening in metabolic 12 engineering 13,14 . 13 14 Synthetic circuits can be implemented at different layers of biological information 15 processing, such as: (i) the genetic layer comprising transcription 15 and translation 16 , (ii) 16 the metabolic layer comprising enzymes 17,18 , and (iii) the signal transduction layer 17 comprising small molecules and their receptors 19,20 . Most designs implemented thus far 18 have focused on the genetic layer, developing circuits that perform computations using 19 elements such as feedback control 21 , memory systems 22,23 , amplifiers 24,25 , toehold 20 switches 26 , or CRISPR machinery 27,28 . However, gene expression regulation is not the 21 only way through which cells naturally perform computation. In nature, cells carry out 22 parts of their computation through metabolism, receiving multiple signals and distributing 23 information fluxes to metabolic, signaling, and regulatory pathways 17,29,30 . Integrating 24 metabolism into synthetic circuit design can expand the range of input signals and 2...