In view of the extremely large number of potential probe sequences to consider, many approaches to microarray design use greedy search or filtering instead of true optimization across the parameter space. Similarly, heuristic shortcuts are employed in lieu of more elaborate thermodynamic models of the probe binding process.
In Leparc et al. (2009), we have shown the improvements to specificity, sensitivity, and array uniformity that can be achieved by non-greedy set-based optimization of microarray probe design. Furthermore, we have introduced an integrated quantitative model of the effects of labelling, cross-hybridization, and other probe and target structures competing for hybridization.
In Mückstein et al. (2010), we have shown the benefits of an improved model for microarray hybridization. Remarkably, specific and unspecific hybridization were apparently driven by different energetic contributions: For unspecific hybridization, the probe–target melting temperature Tm was the best predictor of signal variation. For specific hybridization, however, the effective interaction energy that also considered alternative competing conformations was twice as powerful a predictor of probe signal variation, highlighting the importance of secondary structures in the probe and target molecules.