Integration of clinical chemistry, expression, and metabolite data leads to better toxicological class separation

A large number of databases are currently being implemented within toxicology aiming to integrate diverse biological data, such as clinical chemistry, expression, and other types of data. However, for these endeavors to be successful, tools for integration, visualization, and interpretation are needed. This paper presents a method for data integration using a hierarchical model based on either principal component analysis or partial least squares discriminant analysis of clinical chemistry, expression, and nuclear magnetic resonance data using a toxicological study as case. The study includes the three toxicants alpha-naphthyl-isothiocyanate, dimethylnitrosamine, and N-methylformamide administered to rats. Improved predictive ability of the different classes is seen, suggesting that this approach is a suitable method for data integration and visualization of biological data. Furthermore, the method allows for correlation of biological parameters between the different data types, which could lead to an improvement in biological interpretation.