Assessing performance and tradeoffs of bioforensic signature systems

Chemical and biological forensic programs rely on laboratory measurements to determine how a threat agent may have been produced. In addition to laboratory analyses, it may also be useful to identify institutions where the same threat agent has been produced by the same (or a similar) process, since the producer of the agent may have learned methods at a university or similar institution. In this paper, we evaluate a Bayesian network that combines the results of laboratory measurements with evidence from scientific literature to probabilistically rank institutions that have published papers on the agent of interest. We apply techniques from multiattribute decision science to assess and compare the performance of various implementations of the Bayesian network in terms of three attributes: fidelity, consumption of the forensic sample, and document curation intensity. The mathematical approach we use to compare the various implementations is generalizable to the evaluation of other signature systems.