Pattern-Based Recognition Systems: Overcoming the Problem of Mixtures

The transformative potential of pattern-based sensing techniques is often hampered by their difficulty in dealing with mixtures of analytes, a drawback that severely limits the applications of this sensing approach (the "problem of mixtures"). We show here that this is not an intrinsic limitation of the pattern sensing method. Indeed, we developed general guidelines for the design of the sensing, signal detection, and data interpretation methods to avoid this constraint, which resulted in chemical fingerprinting systems capable of recognizing unknown mixtures of analytes in a single experiment, without separation or pre-treatment before data acquisition. In support of these design principles, we report their successful application to an important analytical problem, metal ion discrimination and quantitation, by constructing a sensor array that provided a linear colorimetric response over a wide range of analyte concentrations. The resulting data set was interpreted using common multivariate data processing algorithms to achieve quantitative identification and concentration determination for pure and mixture samples, with excellent predictive ability on unknowns. Separation and detection methods for analyte mixtures, normally envisioned as independent processes, were successfully integrated in a single system.