Introduction: Artificial Intelligence (AI) and Machine Learning (ML) are currently integrated with pharmaceutical analysis techniques used for drug identification and separation. The integration of analytical techniques such as spectroscopy, chromatography, and mass spectrometry significantly increases sensitivity, specificity, and processing speed. In this review, AI/ML utilization in the identification of different pharmaceutical compounds is presented, including enhancement of performance, classification techniques, and current limitations in pharmaceutical, clinical, and environmental applications. Methods: An extensive review of literature was conducted, reviewing peer-reviewed journals and newer advancements involving AI/ML in pharmaceutical analytical techniques. The considered models are Support Vector Machines (SVM), Convolutional Neural Networks (CNN), decision trees, and ensemble methods. Studies were categorized based on different drug types to identify challenges related to chemical structures, detection, and comparisons to traditional methods. Results and Discussion: The AI/ML models produced have improved ability to detect lowconcentration analytes in complex matrices such as blood, milk, wastewater, and pharmaceutical products. These techniques significantly surpass conventional analytical techniques in detection limits, stability, and computation speed. AI-based spectroscopic data supports rapid and accurate classification, especially in multi-residue analyses. However, model interpretability, data heterogeneity, hardware requirements, and regulation remain major challenges to widespread adoption. Conclusion: Pharmaceutical analytics with the application of AI/ML is of great value for detecting, classifying, and separating compounds, especially in challenging environments. Yet, to reach their maximum capability, issues related to data quality, model interpretability, cross-disciplinary expertise, and regulatory adoption must be addressed. Future research must focus on the development of standardized procedures, enhancing model generalizability, and improving accessibility for broader application in laboratory and field analysis.
Buvaria et al. (Wed,) studied this question.