Physicochemical properties determining drug detection in skin.

Area of Science:

• Pharmacology
• Analytical Chemistry
• Computational Biology

Background:

• Chemicals, including xenobiotics and their metabolites, can be detected noninvasively on skin via swabs and untargeted metabolomics.
• Understanding factors influencing drug presence on skin is crucial for noninvasive monitoring applications.

Purpose of the Study:

• To identify key molecular drivers determining the detectability of systemically administered drugs on the epidermis.
• To develop a predictive model for identifying drugs likely to be found on the skin.

Main Methods:

• Utilized a random forest classifier trained on molecular descriptors of 2561 US Food and Drug Administration (FDA)-approved drugs.
• Employed SHapley Additive exPlanations (SHAP) for model interpretation to identify critical molecular features.
• Evaluated predictive performance using the area under the receiver operating characteristic curve (AUC).

Main Results:

• The random forest classifier achieved a mean AUC of 0.71 in predicting drug detection on the epidermis.
• Analysis identified specific molecular properties influencing drug detectability on the skin.
• Nervous system drugs were predicted as a class more likely to be detected on the skin.

Conclusions:

• Molecular properties can predict the likelihood of detecting systemically administered drugs on the skin.
• Noninvasive skin-based detection of drugs via untargeted metabolomics offers potential clinical benefits.
• This approach could enhance therapeutic drug monitoring, medication adherence assessment, and overall health status evaluation.