DNA microarrays, due to their highly parallel nature, are in principle well suited for rapid identification of known or related microbial species, but our ability to extract meaningful information from microarray images is still at a rudimentary level. The use of DNA microarrays is currently hampered by a few key analytical and theoretical challenges [7, 10]. In particular, the nucleic acid sequence space to be explored can be very large [6], the genetic sequences of many species are very similar, and the concentrations at which the different species are present is typically not known at the time of the sample collection [11], which can result in complex overlapping hybridization patterns. Several analysis methods and experimental designs have been proposed to increase the diagnostic accuracy of identification microarrays. However, there is much disagreement in the literature regarding the merits of particular approaches, as they have been tested on different experimental platforms with samples of varying complexity. Experimental validation of analysis methods is limited, and not feasible as a general strategy [6]. Advances in microarray data analysis would accelerate the employment of the powerful DNA microarray technology, already integrated into lab-on-a-chip instruments [7, 6, 8, 2], in routine clinical practice. For example, the recent discovery of hitherto hidden microbial diversity [1, 2] has led the medical community to recognize the relationship between the microbial communities colonizing the human body and health, disease, and predisposition to disease. There is an increasing awareness of a polymicrobial cause for some diseases, e.g. periodontal disease, rather than attribution to a single causative agent [4]. This could hold the key for explaining the etiology of several hitherto poorly understood diseases (e.g. Chron's disease [5]). However, the characterization of human-associated microbiota is limited by the availability of suitable technologies for rapid microbial identification. We propose to improve the diagnostic accuracy of microarrays and characterize their detection limits by utilizing computational microarray modeling as a tool for design and validation of microarray data analysis methods. For model validation, we will collect thermal hybridization and dissociation data from a novel microfluidic microarray imaging platform (developed in collaboration between Stahl and Yager group). The primary project goal will be achieved through the following specific aims: 1) development of a finite element mathematical model of microarray hybridization that captures the essential features of our platform (competitive binding and dissociation in three-dimensional gel elements, diffusion and convective flow, effects of target length, concentration, and temperature);2) collecting novel thermal hybridization and dissociation data using the integrated microfluidic platform to validate the model in Aim 1;3) using the model to generate simulated datasets, and assess the performance of selected analysis algorithms on datasets corresponding to samples of differing complexity.Narrative DNA microarrays are an exciting new technology for genetic screening and diagnosis of disease. However, they have yet to achieve their full clinical potential for evaluating health and disease states associated with complex mixtures of bacteria. This proposal addresses this untapped potential by developing new tools to guide data analysis and improve the diagnostic accuracy of DNA microarrays.
