Fig. 3: Initial barcode design, classification, and multiplexing.

a Mapping the output strand nanopore-sensitive region. The plot shows the absolute change in mean fractional ionic current manifested by a single-nucleotide mutation at different positions along the strand. b Density plot showing the distribution of mean fractional current for each barcode in Set A. Each distribution is composed of ~14500 data points. c To perform classification, a two-second (20,000 data point) window of the output strand capture event signal is extracted, reshaped into a 2D array, and then used as input to a 2D ResNet-18 CNN. The CNN’s output is a barcode prediction. d Confusion matrix showing classification results of CNN inference on a barcode test set, which achieved an average single-molecule accuracy of 72%. e DSD reaction kinetics plot of two different circuits (barcoded with A0 or A5) multiplexed on the nanopore device. Each gate complex was present at 0.5 uM, fuel strand at 2 uM, input strand at 0.2 uM, and streptavidin at 2 uM. Two samples were prepared: No Input and With Input. The left plot shows the detection of Barcode A0 output for both samples. The right plot shows the detection of Barcode A5 output for both samples.