Optimizing cross-reactivity with evolutionary search for sensors

We report a straightforward evolutionary procedure to build an optimal sensor array from a pool of DNA sequences oriented toward three-way junctions. The individual sensors were mined from this pool under separate selection pressures to interact with four steroids, while allowing cross-reactivity, in a manner designed to achieve perfect classification of individual steroids. The resulting sensor array had three sensors and displayed discriminatory capacity between steroid classes over full ranges of concentrations. We propose that similar protocols can be used whenever we have two or more classes of samples, with individual classes being defined through gross differences in ratios of dominant families of responsive components.