Analytical Data Review on an Artificial Intelligence Platform for Doping Control in Horse Racing

Area of Science:

• Analytical Chemistry
• Forensic Science
• Biotechnology

Background:

• Doping control in horse racing generates vast numbers of chromatograms from GC/MS and LC/MS analyses.
• Manual review of these chromatograms for prohibited substances (PS) is labor-intensive and time-consuming.
• Artificial Intelligence (AI) offers potential for automating image-based data analysis, including chromatograms.

Purpose of the Study:

• To explore the feasibility of using AI for the initial analysis of chromatograms in horse doping control.
• To develop and evaluate an AI model for classifying chromatograms as positive (POS) or negative (NEG) for PS.
• To assess the potential of AI to enhance the efficiency and accuracy of doping control data vetting.

Main Methods:

• Developed a predictive model using Alteryx Designer's image recognition tool.
• Trained the model on over 6000 manually classified chromatograms (POS/NEG).
• Evaluated model accuracy on approximately 700 manually classified chromatograms, achieving over 90% prediction accuracy.

Main Results:

• The AI model accurately identified suspicious/positive and negative chromatograms with no false negatives across two screening methods.
• The model demonstrated high accuracy in classifying chromatograms from LC/MS analysis of horse urine.
• Application to screening methods covering over 300 drug targets showed reliable performance.

Conclusions:

• AI holds significant potential for improving the efficiency of initial chromatogram analysis in horse doping control.
• The developed AI model demonstrates high accuracy and no false negatives, aiding in faster data vetting.
• Challenges remain in reducing analysis time and increasing workflow flexibility for widespread implementation in routine testing.