Sydney Akapame was Awarded a Kopriva Graduate Student Fellowship, 2013-2014
Description of Research
Experimentation is indispensable to the scientific method. Advances in biology, chemistry, pharmacology, pharmacokinetics, and medicine have been predicated on the implementation of experimental designs. Key advances resulted from fitting models based on the experimental data, and specifically fitting models known as nonlinear statistical models. In Sydney Akapame's research, the focus in on use of genetic algorithms (GAs) which is a novel approach to the design problem for nonlinear models.
Sydney's research has wide scientific applicability, and specifically in the pharmaceutical and medical sciences. One research application is calibration for logistic models, which are often used in pharmaceutical applications such as bioassays. A second application is the precise estimation of the absorption rate and elimination rate parameters of compartmental models which are used in the mathematical modeling of how drugs (or other substances) circulate through the body over time. Another application is with the exponential decay model which is widely used in pharmacokinetics and for which estimation of the model parameters and model-based prediction are important experimental objectives. Thus, improved experimentation related to biomedical applications may benefit greatly from this research.
Sydney's current research focuses on the development and implementation of a new design assessment criterion for nonlinear models. Preliminary investigations indicate that GA-based designs based are highly efficient and robust to a wide range of assumptions. In addition, he has made progress in the development of new genetic operators to reduce the GA's computing time in finding optimal designs for nonlinear models. Future research includes the incorporation of weights based on a priori information which is consistent within the scientific (e.g, biological) context of the experiment. Sydney will also explore the potential of combining GAs with existing algorithmic approaches to further improve computing efficiency in design generation for nonlinear models..