Extended Data Fig. 5: Characterization of the learning motif.
a, Sequence-level diagrams of inhibited activator \(Ac{t}_{1,1}^{* }\), input X1, and label L1. The \({5}^{{\prime} }\) end of the bottom strand in the inhibited activator is modified with a fluorophore, and the \({3}^{{\prime} }\) end of the input strand is modified with a quencher. The double-stranded activator produced from this reaction has the quencher adjacent to the fluorophore, resulting in decreased fluorescence, which is then normalized to increased activator concentration based on control experiments. b, Sequence-level diagrams of inhibited activator \(Ac{t}_{1,2}^{* }\), input X1, and label L2. c, Sequence-level diagrams of inhibited activators \(Ac{t}_{5,1}^{* }\) and \(Ac{t}_{5,2}^{* }\), along with input X5. Label strands are the same as shown in panels a and b. d, Fluorescence kinetics experiments and simulations with distinct learning gates and input-label pairs. Fluorescence kinetics data (dotted trajectories) are overlaid with mass-action simulations of chemical kinetics by solving ordinary differential equations (solid trajectories). Details of modeling see Supplementary Note 2.2. The small variations of reaction kinetics across distinct learning gates can be explained by the sequence of the Lj domain and the minor secondary structure within the U* and Xib* domains. 1 × indicates a standard concentration of 50 nM. e, Crosstalk evaluation. Excess label (5×) was used. Four distinct fluorophores, ATTO488, ATTO550, ATTO590, and ATTO647 were used on \(Ac{t}_{1,1}^{* }\), \(Ac{t}_{3,1}^{* }\), \(Ac{t}_{5,1}^{* }\), and \(Ac{t}_{7,1}^{* }\), respectively. The same set of fluorophores with a different order were used on the other four inhibited activators. Eight hours of fluorescence kinetics data are shown in Fig. 2i whereas endpoint data of activator concentrations (relative to the standard concentration) at 8 hours are shown here. f, Reversibility evaluation. Before 16 hours, all fluorescence kinetics trajectories are repeats of the same sample with 1.5 × learning gate (\(Ac{t}_{5,1}^{* }\) or \(Ac{t}_{5,2}^{* }\)), 0.5 × input (X5), and 5 × label (L1 or L2). Label inhibitor (Inh1 or Inh2) with varying concentrations was added at 16 hours and fluorescence data was collected until 32 hours. No fluorescence decrease was observed with increased inhibitor concentration, suggesting robust irreversibility of the learning gate.
