Statistical analysis and modeling of bioprocesses

Abstract

This diploma thesis deals with the use of statistical methods and mathematical modeling of biological processes. Due to the complexity of biological systems and high amount of variables, statistical and numerical methods for parameter prediction are required.<br />A bioprocess of Escherichia coli producing recombinant protein was analyzed. The variable of interest which we predicted with the help of different statistical methods was \textit{production rate}. In the first part of this work we applied PLS (Partial Least Squares Regression) on our data. Furthermore, we used Neural Networks, built a model with the data of the one cultivation and, in order to validate the model, applied it on the data of another cultivation. The Neural network model was applied on the whole and on the by PCA (Principal Component Analysis) reduced data set.<br />In the next part of this thesis statistical methods were combined with numerical methods in order to achieve better prediction of the variable of interest. Chemical kinetics of the analyzed bioprocess were described with the help of differential equations. Their parameters were estimated by use of neural networks. By achieving that very promising results were obtained. In the last part of our work, we applied Bayesian statistical methods for variables selection. In this way we were able to find key parameters of observed biological process.<br />Finally,the obtained results were compared and future steps based on the obtained results were proposed.