Chemometric calibration for Raman spectral analysis - Machine learning and hard modeling in CHO cell culture applications
Loading...
URL
Journal Title
Journal ISSN
Volume Title
Perustieteiden korkeakoulu |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
2023-12-12
Department
Major/Subject
Systems and Operations Research
Mcode
SCI3055
Degree programme
Master’s Programme in Mathematics and Operations Research
Language
en
Pages
54+11
Series
Abstract
In this thesis, support vector regression (SVR), science-based calibration (SBC), and partial least square (PLS) regression models are constructed for lactate, ammonia, and amino acid concentration in Chinese hamster ovary (CHO) cell cultures. To construct and test said models, data from five different CHO perfusion bioreactor runs was obtained, consisting of time-gated Raman spectra and the concentrations of the analytes of interest that coincide with the given spectra. In addition to the data from the bioreactor runs, response spectra of the various amino acids and of the bioreactor media were obtained for conducting SBC. It was found that both SVR and PLS models produce predictions of satisfactory accuracy for roughly half of the analytes, with the best results occurring for amino acids. When comparing the results of the SVR and PLS models, we saw that in nearly every case, SVR can produce more accurate predictions than PLS, especially when non-linear forms of SVR are utilized. When it comes to SBC, we were unable to produce models with adequate accuracy, due to a failure to accurately scale the response spectra of the various analytes to represent the change in intensity per unit concentration. As far as the need for reference data, although the results were not satisfactory, the processes involved with SBC required a fraction of the reference data that PLS and SVR did.Description
Supervisor
Hakula, HarriThesis advisor
Daniel, AmuthachelviKeywords
chemometric calibration, support vector regression, supervised machine learning, science-based calibration, bioprocessing, Raman spectroscopy