Abstract:
Current biosensors are primarily based upon previous observations: they detect
organisms known to be pathogenic. Future biowarfare agents, however, are likely to
contain completely novel or re-engineered proteins and nucleic acid sequences intended
either to make previously harmless organisms pathogenic, to increase the pathenogenicity
of existing agents or expressly to render the agent undetectable by conventional serotype-or
PCR-based methods. The present work describes the creation and validation of a
nucleic-acid microarray-based biosensor for the detection of putative biohazards present
in environmental air samples. The prototype array consists of 15,200 pseudo-random
25bp oligonucleotide probes whose sequences were generated using variable-length
Markov chain models trained on sequence from pathogenic prokaryotic genomes.
Classifiers constructed on organism-specific patterns of hybridization were then applied
to unknown or mixed samples to determine a likelihood of detection. With this approach,
the ability to estimate the presence of a novel or engineered threat then requires only the
characterization of the binding pattern of the agent’s amplified genomic DNA to the
array.
