Application of multivariate analysis for optimizing & predicting hepatic function

Abstract

An optimization model based upon multivariate analysis was developed to capture hepatic specific function in relation to the environmental condition and the intracellular metabolic network and the flux information obtained from metabolic flux analysis (MFA). Fisher discriminant analysis (FDA) was applied to maximize the discrimination among groups thus permitting visualization of the sample separation between different conditions. FDA identified factors that contribute greatly to the separation of the groups. Mapping fluxes to a hepatic function permits an examination of the interrelationship of the fluxes and captures the hepatic function in terms of the metabolic profile. Partial least square (PLS) was the mapping technique applied to evaluate the effect of metabolic state on hepatic function, namely, the levels of intracellular triglyceride or urea production. This methodology identified fluxes most relevant to minimizing the accumulation of intracellular triglyceride and maximizing the production of urea, two important hepatic functions. In the study, 75 metabolic fluxes were mapped to measured levels of intracellular triglyceride or urea. Once a mapping model was constructed, analyzing the model parameters permitted the assessment of how the metabolic profile, in turn, pathways collectively regulate and control hepatic function by identifying pathways that are highly correlated with the hepatic function.