Applications of image analysis in the development of recombinant protein processes
Recombinant proteins are produced using a variety of fermentation, cell culture and transgenic animal processes. Because fermentation and cell culture play such a prevalent role in protein production, it is important that scientists use every tool to gain as much understanding of these processes as is possible. One tool that is utilized in almost every fermentation and cell culture laboratory is the optical microscope. Biologists and engineers alike use the microscope to gather detailed information about fermentation and cell culture processes. Unfortunately, much of the information provided by the microscope is qualitative in nature, and can be lost when technical transfers are made between scientists. Although photographs assist in retaining this qualitative information, it is difficult to systematically utilize this information. Image analysis can provide systematic information contained in images that would be difficult or impossible for scientists to duplicate manually. In this thesis, two applications of image analysis were developed to demonstrate the role that image analysis can play in the development of recombinant protein processes. ^ In the first application, a semi-automated image analysis method was developed for the determination of population characteristics during recombinant Escherichia coli fermentations. Identification of objects within an image was performed by binarization of enhanced images. Several factors related to size, shape, intensity, and pixel distribution were calculated for each image. These variables were first used to identify bacterial and non-bacterial objects. Then the bacterial objects were classified using a probabilistic neural network. Finally, the probabilistic neural network allowed bacteria to be classified according to their growth characteristics and whether they contained inclusion bodies. ^ In the second image analysis application, a semi-automated image analysis method was developed for the determination of population characteristics of Chinese Hamster Ovary cells during recombinant protein runs. Binarization of enhanced images allowed identification of objects within an image. Factors related to size, shape, intensity, and pixel distribution were calculated for each object. Principal component analysis of these factors was utilized to develop a modified probabilistic neural network for object classification. After classification and filtering out non-cellular objects, cell populations were categorized. Cell counts and viability were estimated, and cells were classified as single cells, doublets or cellular aggregates. A comparison with manual observations is presented. ^
Roderick William Geldart,
"Applications of image analysis in the development of recombinant protein processes"
Dissertations and Master's Theses (Campus Access).