Identification of substandard and falsified vaccines using spectroscopic techniques and machine learning algorithms

Watson, M (2025) Identification of substandard and falsified vaccines using spectroscopic techniques and machine learning algorithms. Other thesis, Liverpool John Moores University.

Preview

Text 2025WatsonMPhil.pdf - Published Version Available under License Creative Commons Attribution Non-commercial. Download (6MB) | Preview

Abstract

Covid-19 is a novel coronavirus first noted in 2019. Accumulating over 775 million cases and 7.1 million deaths worldwide, vaccination was presented as the most effective solution for Covid-19 related harm reduction. However, the accelerated development and limited availability of Covid-19 vaccines opened a gap in the market for the emergence of substandard and falsified (SF) vaccines. SF Covid-19 vaccines were seized around the world, with impacts on public health and the economy. Traditional analytical techniques are costly and require specialist training, as well as sophisticated laboratory environments.
This thesis evaluated the application of vibrational spectroscopic techniques, specifically ATR-FTIR, Raman, and SERS, for the rapid, cost-effective, and robust analysis of Covid-19 vaccines and the identification of SF vaccines. The study focused on the use of both laboratory-based and portable handheld instruments, assessing their potential for both laboratory and field-based vaccine authentication.
The ATR-FTIR method was optimised using a Perkin Elmer Spectrum Two FTIR spectrometer with an ATR diamond accessory. It was applied to Covid-19 vaccines, comprising DNA-based, mRNA-based, and live attenuated formulations. The method was effective in characterising vaccine brands, yielding high-quality spectra and achieving excellent results in vaccine identification. Key spectral features, including nucleic acid-specific bands, were identified, allowing for clear discrimination between vaccine brands. A classification model, specifically KNN, achieved 99.7% accuracy in classifying vaccine brands, demonstrating ATR-FTIR’s potential for semi-automated vaccine authentication.
Raman and SERS methods were optimised using the Metrohm MIRA XTR DS handheld Raman spectrometer. Despite positive results in vaccine characterisation, the performance of SERS was variable, with lower enhancement factors observed in the final study compared to the pilot study. Vaccine degradation over time and changes in colloidal silver substrates were cited as contributing factors. Nonetheless, classification models achieved 100% accuracy, precision, and recall, indicating strong potential for SF vaccine detection with continued method development.
Raman microscopy, performed using the Horiba XPloRA™ PLUS, demonstrated excellent potential for Covid-19 vaccine identification, particularly in samples with minimal fluorescence interference. However, the presence of fluorescence in most spectra limited the method’s reliability, highlighting the need for further method optimisation and larger datasets.
Overall, ATR-FTIR was identified as the most robust and effective technique for rapid, non-destructive vaccine analysis, while Raman and SERS showed promise but required further refinement. This research supports the use of ATR-FTIR combined with machine learning as a powerful tool for Covid-19 vaccine authentication and the detection of SF vaccines.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Vaccine; Covid-19; Spectroscopy; Machine Learning; ATR-FTIR; Raman spectroscopy; Raman microscopy; Substandard; Falsified
Subjects:	R Medicine > R Medicine (General)
Divisions:	Pharmacy and Biomolecular Sciences
Date of acceptance:	15 May 2025
Date of first compliant Open Access:	25 June 2025
Date Deposited:	25 Jun 2025 12:51
Last Modified:	25 Jun 2025 12:52
DOI or ID number:	10.24377/LJMU.t.00026633
Supervisors:	Assi, S, Birkett, J, Al-Jumeily OBE, D and Khan, I
URI:	https://researchonline.ljmu.ac.uk/id/eprint/26633

View Item