Automated Microarray Organism Detection with a Non-Gaussian Maximum Likelihood Model

The utility of DNA microarrays for bioagent detection and classification will only be fully realized when hybridization intensities can be accurately related to sequence and abundances of constituent DNA molecular fragments in the sample. To move toward this goal, we have developed a procedure that is robust to suboptimal image quality and a maximum likelihood-based processor to estimate the concentration of bioagent targets in a reaction. The signal models used for the maximum likelihood processing are based on the physics of DNA microarray hybridization. An adaptive background signal model was included to manage the wide variation of background clutter expected in a typical bioagent detection scenario.