Development of a screening tool to detect identical urine samples within the athlete biological passport based on artificial intelligence

Code: 22D06MT

Several cases of so-called sample swapping have been detected in recent years. During sample swapping an athlete substitutes the presumable positive doping control sample by another negative urine sample in order to circumvent an adverse analytical finding. For the exchange of urine, a sample from the same individual or from a different individual may be used. While the latter can result in a significantly different steroid profile, employing a urine sample from the same individual collected prior to administration of a doping agent will not show any differences in the steroid profile. Here it can be expected that the stored clean sample will be used several times and that this may result in steroid profiles with an atypical similarity. Two cases of this type of sample swapping have already been detected, but only by chance as the similar samples were analyzed in the same laboratory in a timely manner. 
The aim of this research project is the development of a screening tool based on machine learning and artificial intelligence which will enable to detect steroid profiles with an atypical similarity in all steroid profiles determined world-wide in different accredited doping control laboratories. The developed software will be able to search within the existing profiles for similar ones and will be trained to compared new entries to the dataset to existing ones in order to indentify highly similar patterns. Employing artificial intelligence will enable to train the software to become highly sensitive, selective and robust. In order to confirm potentially similar samples to belong to the same athlete, DNA analysis of these urine samples will be applied where applicable.