An Expert System for the Pattern Recognition of Biological Effects

The objective of this project is to develop an expert system capable of quantifying and interpreting complex, time-related, global changes in gene, protein expression, and metabolite pattern profiles. The expert system will identify toxic effects with exposure at very low, sub-symptomatic concentrations to military personnel. The project will meet five challenges to the development of the system: (a) Identify all specific biomarkers of a database of model toxins (b) Predict the specific impact sites of the toxin effected organs (i.e. mechanisms of toxicity) by pattern classification of characteristic biomarkers and predict the extent of damage compared to the response to the most relevant model toxins (equivalent doses); (c) Quantify the dose response curve especially the threshold level with estimated probabilistic confidence; (d) Predict the time evolution of the biological effects in genes, proteins, and metabolites using available time-dependent measurements of all pattern profiles; (e) Integrate the analyses of all available pattern profiles and optimize the synergized capability of interpreting and predicting the biological effects. In Phase I, we have developed accurate multivariate statistical algorithms for identifying gene and metabolite biomarkers and predicting the dose levels and time points of unknown samples. The feasibility and extendibility of the concept in terms of accuracy, efficiency, and applicability have been demonstrated successfully with practical examples. Phase II work will aim at expanding the capability of the expert system in an integrated systems biology approach.