This Phase II proposal continues the development and validation of a bioinformatics and diagnostic platform for rapid detection and pre-symptomatic identification of biological agents on the basis of host gene and proteomic expression response profiles. Our collaborators preliminary host-pathogen studies indicate that exposed individuals display idiosyncratic peripheral blood mononuclear cell (PBMC) mKNA expression and serum protein temporal patterns to pathogenic agents prior to the onset of full illness. New tools are needed to determine the patterns from the volumes of complex, variable, and noisy genomic/proteomic data generated from host-pathogen studies. Our computational tools are based on the probabilistic power of dynamic Bayesian networks (DBNs) which are utilized to learn, model and recognize the dynamic pattern-of-change of mRNA and proteins (""""""""biosignature"""""""") of the hostpathogen innate immune response. A unique feature of our approach is the inclusion of """"""""time"""""""" combined with prior quantitative and qualitative knowledge that improves the recognition accuracy between different pathogenic agents. Phase I s/w prototype demonstrated proof-of-concept that our computational approach produced correct DBN models from time-course data mimicking the innate immune response to nine different pathogen types. The prototype performed remarkably well in both the representation of the time-course biosignatures as DBN models and for correctly identifying an unknown biosignature to the correct infectious agent with better than 98% accuracy. Phase II main objective is to statistically show that Seralogix's computational framework can extract and model unique host-pathogen biosignature patterns in the face of host heterogeneities using real time-course PBMC mRNA and protein datasets for the pathogens/toxin B. anthracis, cowpow, and staphylococcal enterotoxin B. in animal and non-human primate models. Existing time course data will be provided by our collaborators at the Walter Reed Army Institute of Research and the Biosignature Consortium (comprised of the University of New Mexico, the University of Texas Southwestern Med. Ctr., and Lawrence Livermore Labs). Seralogix believes that our DBN based computational tools will be important for: 1) deciphering the cellular signaling pathways and mechanisms of virulence and toxicity of pathogens/toxins; 2) creating new diagnostics for real-time, pre-symptomatic pathogenic identification, and 3) understanding the progressive stages of a disease to aid in creating new intervening drugs and therapeutic strategies. ? ? ? ?
