Novel Algorithms for Comprehensive Untargeted Detection of Doping Agents in Biological Samples.

Abstract

To address the limitations of current targeted analytical methods that can only detect known doping agents, a novel methodology that permits untargeted drug detection (UDD) has been developed to help in the fight against doping in sports. Fifty-seven drugs were spiked into blank equine plasma and were treated as unknowns since their exact masses and chromatographic retention times were not utilized for detection. The spiked drugs were extracted from the plasma samples and were analyzed using liquid chromatography coupled to high-resolution mass spectrometry (LC-HRMS). The acquired LC-HRMS raw data files were processed using metabolomic software for compound detection and identification. For UDD with the resultant data, a mathematical model was created, and two algorithms were generated to calculate the ratio of the mean (ROM) and outlier index (OLI). Using ROM and OLI, the majority of the 57 drugs were accurately detected by name (52 of 57) or chemical formula (1 of 57). The limit of detection for the drugs was from tens of picograms to nanograms per milliliter. Xenobiotics and endogenous substances relevant to doping control were also identified using this untargeted approach following their extraction from real-world race samples, thus validating the UDD methodology. To the authors' knowledge, this is the first completely UDD methodological approach and represents significant advance toward using artificial intelligence for the detection of both known and emerging doping agents in sports.