An investigation of preprocessing and parameter estimation methods for hit selection and dynamical modelling of genetic circuits in a systems biology context

• Main Authors: Yu-HengWu, 吳雨衡
• Other Authors: Torbjörn Nordling
• Format: Others
• Language: en_US
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/ym53q2

碩士 === 國立成功大學 === 機械工程學系 === 107 === Systems biology studies the interaction of biological components from a systematic point of view, which benefits applications, such as synthetic biology and drug discovery. The field took off thanks to high-throughput measurement technologies in the late 1990s. To extract useful information from the measurements, it is crucial to do system identification and data preprocessing. Here we investigate data preprocessing and parameter estimation methods on two separate topics of systems biology, which are system identification of the GAL1 synthetic genetic circuit from fluorescence microscopy data and image analysis for hit selection based on protein microarray data. Modelling of GAL1 genetic circuit in yeast using three equations Background: Synthetic genetic circuits can be used to modify and control existing biological processes. Currently, their use is largely hampered by the trial and error approach used to design them. Lack of reliable quantitative dynamical models of genetic circuits obstructs the use of well-established control design methods. Aim: We investigate the GAL1 synthetic genetic circuit as a first step toward the creation of a pipeline for automated identification of synthetic genetic circuits. Method: We study modelling from the system identification perspective on the yGIL337 strain of S. cerevisiae. In the strain, expression of a fluorescent reporter can be turned on by growing the yeast in galactose and off by growing it in glucose. We estimate the parameters of our three ordinary differential equations (ODE) of Michaelis-Menten type based on published data from an in vivo microfluidic experiment after redoing the data preprocessing. Results and conclusion: We show that the goodness-of-fit of our three ODE model is comparable to five previously proposed models and hypothesize that the system is an adaptive feedback system. We also show that the data is not informative enough to invalidate any of the alternative models despite significant difference in their structure. Analysis of data preprocessing and systematic errors in protein microarray for hit se- lection Background: Protein microarrays allow rapid testing of molecular binding of thousands of proteins on a single chip. However, in practice, the microarray chip contains artefacts due to inevitable experiment errors making hit selection challenging. Aim: We aim to develop an automatic pipeline for hit selection optimised for protein microarrays, including data preprocessing. In this project, we focused on image preprocessing to detect, quantify, and exclude artefacts of protein microarrays as the first step. Method: Center finding, spot segmentation, background surface fitting, and smear surface fitting are implemented to remove systematic errors. Results and conclusion: We optimise the identification of protein spots, background intensities, and smear intensities in the protein microarray image. A 5th order polynomial surface fitting is then applied on background and smear data, respectively. The goodness-of-fit of background and smear give R2 equal to 0.559 and 0.488.