The impressive capabilities of the mammalian brain—ranging from perception, pattern recognition and memory formation to decision making and motor activity control—have inspired their re-creation in a wide range of artificial intelligence systems for applications such as face recognition, anomaly detection, medical diagnosis and robotic vehicle control1. Yet before neuron-based brains evolved, complex biomolecular circuits provided individual cells with the ‘intelligent’ behaviour required for survival2. However, the study of how molecules can ‘think’ has not produced an equal variety of computational models and applications of artificial chemical systems. Although biomolecular systems have been hypothesized to carry out neural-network-like computations in vivo3,2,4 and the synthesis of artificial chemical analogues has been proposed theoretically5,6,7,8,9, experimental work10,11,12,13 has so far fallen short of fully implementing even a single neuron. Here, building on the richness of DNA computing14 and strand displacement circuitry15, we show how molecular systems can exhibit autonomous brain-like behaviours. Using a simple DNA gate architecture16 that allows experimental scale-up of multilayer digital circuits17, we systematically transform arbitrary linear threshold circuits18 (an artificial neural network model) into DNA strand displacement cascades that function as small neural networks. Our approach even allows us to implement a Hopfield associative memory19 with four fully connected artificial neurons that, after training in silico, remembers four single-stranded DNA patterns and recalls the most similar one when presented with an incomplete pattern. Our results suggest that DNA strand displacement cascades could be used to endow autonomous chemical systems with the capability of recognizing patterns of molecular events, making decisions and responding to the environment.
Access optionsAccess options
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
We thank P. Rothemund, P. Yin, D. Woods, D. Soloveichik and N. Dabby for comments on the manuscript. We also thank R. Murray for the use of experimental facilities. This work was supported by the NSF (grant nos 0728703 and 0832824 (The Molecular Programming Project)) and by HFSP award no. RGY0074/2006-C.
This file contains Supplementary Figure 1-28 with legends, Supplementary Methods, Supplementary Text and Data comprising: 1 The seesaw DNA gate motif; 2 Four types of seesaw gates; 3 Four transformation rules; 4 Circuit design lessons learned from experiments; 5 In silico training of dual-rail monotone Hopfield associative memories; 6 A four-neuron dual-rail monotone Hopfield associative memory; 7 Modeling and simulations; 8 Sequences, Supplementary Tables 1-7 and additional references.