Convolutional Neural Networks Assisted Peak Classification in Targeted LC-HRMS/MS for Equine Doping Control Screening...

Area of Science:

• Analytical Chemistry
• Computational Biology
• Veterinary Science

Background:

• Doping control screening relies on time-consuming manual analysis of chromatograms.
• High-throughput screening of numerous compounds and metabolites is essential in modern anti-doping efforts.
• Machine learning, particularly convolutional neural networks (CNNs), offers potential for automating analytical tasks.

Purpose of the Study:

• To evaluate the feasibility and accuracy of deep learning for automating equine doping control.
• To develop an AI-driven strategy for chromatogram classification in biotherapeutics screening.
• To address the challenge of achieving a zero false negative rate (FNR) in automated doping analysis.

Main Methods:

• Implementation of a deep learning strategy using CNNs for chromatogram image classification.
• Integration of a linear discriminant analysis (LDA) classifier.
• Training and validation using data from ultra-high-pressure liquid chromatography coupled to high-resolution tandem mass spectrometry (UHPLC-HRMS/MS).

Main Results:

• Demonstrated the feasibility and accuracy of a deep learning approach for equine doping control.
• Developed a CNN scoring model combined with an LDA classifier for robust analysis.
• Achieved high accuracy in classifying chromatograms, essential for reliable doping control.

Conclusions:

• Deep learning strategies are effective for automating chromatogram classification in equine doping control.
• AI, specifically CNNs and LDA, can significantly enhance the efficiency and accuracy of doping control laboratories.
• The proposed AI tool shows promise for future applications in anti-doping efforts, ensuring zero FNR.