SVM-Assisted FTIR Spectroscopic Analysis of Ampicillin and Amoxicillin under Degradation Conditions
Vinicius Anjos1, Daniella Pereira1, Rafael De Góes2, Denise Zezell1; 1University of São Paulo, São Paulo, Brazil; Federal University of Technology - Paraná, Curitiba, Brazil.
Abstract
The quality of drugs has long been a concern for the World Health Organization (WHO), as no health service can function effectively without ensuring that these products meet established standards of quality, safety, and efficacy. Machine learning models and Fourier Transform Infrared (FTIR) spectroscopy have become powerful tools for improving antibiotic quality control, thanks to their ability to analyze complex sample datasets derived from chemical and genomic profiles. This study explores the application of FTIR spectroscopy combined with Support Vector Machine (SVM) classification to monitor the degradation of important antibiotics, Ampicillin and Amoxicillin. Wasserstein Generative Adversarial Networks (WGANs) were employed for data augmentation. Attenuated Total Reflection (ATR) – FTIR was performed to investigate these antibiotics over different degradation processes. Performance metrics demonstrated strong classification capabilities; the model effectively distinguished between intact and thermally degraded Ampicillin and Amoxicillin samples, achieving an average accuracy of approximately 89.8% with low variance across folds. All analyses were performed in OriginLab and Python. WGAN augmented spectra improved SVM performance, demonstrating the impact of its use when the amount of real data available is limited. Our results highlight the potential of integrating FTIR spectroscopy with advanced machine learning architectures to provide a rapid, non-destructive, and reliable method for antibiotic quality control under degradation conditions.
Speaker
Vinicius Anjos
University of São Paulo
Brasil
Discussion
Ask question
