|Title:||Data Science Methods to Decrease Experimental Efforts in Quality by Design Tasks||Language:||English||Authors:||Borchert, Daniel||Qualification level:||Doctoral||Keywords:||Quality by Design; Data Science; Bioprocess Development||Advisor:||Herwig, Christoph||Assisting Advisor:||Spadiut, Oliver||Issue Date:||2020||Number of Pages:||127||Qualification level:||Doctoral||Abstract:||
Digitization in the pharmaceutical industry, Pharma 4.0, is becoming more important every year. The first concepts in this area are already being implemented by several large companies and will be completed in the next few years. Such concepts should make it more comfortable in the future to collect data, exchange knowledge, and conduct holistic data analyses.The expected benefits of such systems are highly appreciated and, according to the current state of the industry, also needed, especially in the Quality by Design (QbD), which is applied to guarantee consistent product quality. QbD is used to understand the source of the variance of particular quality attributes as well as to understand the linkages of its interaction in order to ensure robust product quality. However, it is difficult to apply the widely used and often applied QbD approach correctly because, from a current perspective, there is a lack of basic QbD interpretation and sufficient statistical knowledge of the process experts. Therefore, more and more experiments will often be conducted within the QbD study in order to get the wanted process understanding. However, mechanistic process modeling and data analysis approach should be used to get to this aim.The goal of that thesis is to go back to the roots, such as using process knowledge from models and statistical analysis, by investing more time in data analysis rather than in experimental effort by applying existing data science methods useful within the QbD approach. Furthermore, we aim to show that a combination of existing methods and correct result interpretation lead again to the QbD benefits as initially anticipated. As particular achievements of this thesis, it is demonstrated how a simple combination of existing root cause analysis approaches is used to improve the cause and effect analysis to gather additional and more profound process knowledge using historical data. This method provides new insights into the process by using all available process data within data analysis and useful result interpretation. Since robust product quality is one of the most relevant results of the QbD approach, risk assessment and interpretation is the core element of that concept. Improvements of the current risk assessment and interpretation approaches of the individual process parameters are demonstrated, to finally get a quantitative risk evaluation for each process parameter individually. In the end, a comprehensive case study is presented to demonstrate the implementation and benefits of the improved QbD approach. A holistic design space evaluation at the end of the QbD study is shown and lays the basis for a holistic process control strategy.The overall goal to reduce the number of experiments by investing more time in useful data analysis is shown along with the first implementation example of the newly developed approach. It is shown that appropriate interactions into the current QbD approach have a significant influence on the experimental effort and a high potential to reduce the time to market and overall process costs, to finally reduce the patient risk since drugs can be faster available.
|DOI:||10.34726/hss.2020.81804||Organisation:||E166 - Institut für Verfahrenstechnik, Umwelttechnik und technische Biowissenschaften||Publication Type:||Thesis
|Appears in Collections:||Thesis|
Show full item record
Files in this item:
|Data Science Methods to Decrease Experimental Efforts in Quality by Design Tasks.pdf||638.01 kB||Adobe PDF|
checked on Feb 27, 2021
checked on Feb 27, 2021
Items in reposiTUm are protected by copyright, with all rights reserved, unless otherwise indicated.