Integrating Advanced Metabolomics and Machine Learning for Anti-Doping in Human Athletes.

Area of Science:

• Sports science
• Biochemistry
• Computational biology

Background:

• Doping in sports remains a significant challenge, necessitating advanced detection strategies.
• Traditional doping detection methods primarily serve forensic purposes by identifying prohibited substances directly.
• Metabolomics offers complementary insights into both exogenous compounds and endogenous physiological changes, potentially extending detection beyond direct drug presence.

Purpose of the Study:

• To review the integration of metabolomics with AI and ML for anti-doping efforts.
• To explore how metabolomics can reveal physiological alterations indicative of doping.
• To highlight the role of computational tools in analyzing complex metabolomic data for athlete profiling.

Main Methods:

• Utilizing high-throughput platforms like UHPLC-HRMS and NMR for metabolomic data acquisition.
• Employing targeted and untargeted metabolomic workflows for comprehensive profiling.
• Applying machine learning algorithms (supervised and unsupervised) for pattern recognition, classification, and anomaly detection.

Main Results:

• Metabolomics datasets, when analyzed with ML, can effectively discriminate between doped and clean athlete profiles.
• AI and ML enhance the sensitivity and specificity of doping screening, optimizing resource allocation.
• Case studies demonstrate the detection of doping agents like r-HuEPO and steroids through indirect markers and extended detection windows.

Conclusions:

• Integrating metabolomics and ML provides a powerful, complementary approach to traditional forensic testing in anti-doping.
• These technologies can identify doping through indirect physiological markers and offer extended detection capabilities.
• Future advancements require interdisciplinary collaboration and open data infrastructure to enhance fair play and athlete well-being.