A competitive hybridization model predicts probe signal intensity on high density DNA microarrays

A central, unresolved problem of DNA microarray technology is the interpretation of different signal intensities from multiple probes targeting the same transcript. We propose a competitive hybridization model for DNA microarray hybridization. Our model uses a probe-specific dissociation constant that is computed with current nearest neighbor model and existing parameters, and only four global parameters that are fitted to Affymetrix Latin Square data. This model can successfully predict signal intensities of individual probes, therefore makes it possible to quantify the absolute concentration of targets. Our results offer critical insights into the design and data interpretation of DNA microarrays.