SELECTION OF A MATHEMATICAL MODEL FOR THE KINETICS OF Haemophilus influenzae TYPE B USING AKAIKE’S INFORMATION CRITERION

Abstract

Haemophilus influenzae type b (Hib) vaccine is made up from its capsular polysaccharide (PRP). Low productivity of the polysaccharide during cell growth increases the final cost of this vaccine. Hib achieves low levels of cellular concentration in vitro due to the inhibition caused by acetate. The Akaike Information Criterion (AIC) was used in this work for selecting models of microbial growth. The application to the case of the multivariate models is outlined and the procedure is carried out using data from Hib cultures. From 4 models of biomass and 15 of acetate and PRP, one could be selected with great evidence for support. The use of AIC has shown to be robust and free of subjectivity, and it was able to define a kinetic model that is adequate for the cell growth and the production of its PRP over a wide range of culture conditions. The exponential inhibition factor was found to be the best for modelling inhibition of cell growth by acetate, while the hyperbolic factor was the best for inhibition of PRP formation. The acetate formation was found to have both growth associated and nonassociated types. PRP formation was found to be only growth associated.

Kinetic modelling: Regression and validation stages, a compulsory tandem for kinetic model assessment

“…For instance, the modeller could compare the residual sum of squares, plot the parity plot, or calculate the Akaike information criterion (AIC) to find the most reliable models. [ 15–18 ]…”

“…For instance, the modeller could compare the residual sum of squares, plot the parity plot, or calculate the Akaike information criterion (AIC) to find the most reliable models. [15][16][17][18] The fact that the validation stage is not commonly used could be due to the fact that the kinetic modeller only considers the holdout validation method, leading to the sacrifice of some experimental runs from the regression stage. The cross-validation (CV) approach, not common in kinetic modelling articles, can be used to overcome this issue.…”