Current point sensors for biological weapons detect a finite number of organisms. This limits detector flexibility and requires significant physical reengineering when a novel pathogen is discovered. The current work seeks to design a microarray-based sensor using pseudo-random oligomer probes paired with pattern recognition and classification algorithms. The prototype array consists of 15,000 25bp probes derived from a variable-length Markov chain trained on pathogenic prokaryotic genomic DNA. This serves to increase the probability of binding to prokaryotic genomic sequence while decreasing the incidence of binding to non-prokaryotes over those rates one might expect by chance for purely random 25-mers. A comparison of the performance of various classification algorithms is reported and discussed in the context of identification of the hybridization patterns of simulants B. cereus, B. subtilis and P. agglomerans against a complex, mixed-genome background from outdoor environmental sampling. Detection of an emerging infectious pathogen or engineered organism using such a detector would then require only initial characterization of that organismÕs genomic hybridization pattern on the biosensor array. Fielded sensors can then be upgraded with only a software update pushed out over existing reporting networks rather than by costly physical reengineering.