Detecting Fentanyl in Human Nails and Exploring the Inhibition Mechanism of Fentanyl on Butyrylcholinesterase via Spectroscopy and Machine Learning

Fentanyl, a synthetic, illicitly manufactured opioid, has been the cause of a continuous, steep increase of drug overdose cases across recent decades. Fentanyl is extremely potent, lethal, and dangerous. Traces of it can be found in bodies of water across the world, making it a danger for anyone who may encounter it, not just drug manufactures, dealers, or users. Research into detection methods and biological effects of fentanyl are highly important today. Here, we discuss two different projects related to fentanyl. In one project, we explore a novel detection method using human nails as a sample matrix by using ATR-FTIR spectroscopy together with PLS-DA and SVM-DA machine learning techniques to determine if nails containing fentanyl can be distinguished from those without. In another, ongoing project, we attempt to discover the currently unknown inhibition mechanism of two fentanyl analogues on butyrylcholinesterase through enzymatic assay, Lineweaver-Burk charts, and Raman spectroscopy combined with PCA and PLS machine learning techniques. The results of the first project demonstrate that simple visual examination of spectra is not very helpful when attempting to distinguish between control and fentanyl groups. However, both machine learning models performed extremely well when asked to categorize the data into the two groups. Both models performed with high specificity, sensitivity, and overall accuracy at all stages of model creation, optimization, and validation. This excellent performance only degraded slightly when exposed to unseen data, only further proving their robustness. ROC curves for the PLS-DA model indicated an overall accuracy of 98% for cross validation and an overall accuracy of 94% for external validation. For the SVM-DA model, the ROC curves indicated 97% overall accuracy for cross validation and 89% for external validation. When classification was considered at the donor level—i.e., determining whether the donor of the nail sample was using fentanyl—all donors were correctly classified. The second project discussed here is still ongoing. Firstly, BChE was inhibited with 2 different fentanyl analogues and enzymatic activity was assessed via the Ellman assay method. These results were used to create Lineweaver-Burk charts to determine if a main inhibition mechanism could be observed. Lineweaver-Burk analysis indicated mixed type inhibition for one fentanyl and no inhibition for the other. The Raman spectra of these samples were then collected and processed using PCA and PLS machine learning methods. These results demonstrated that variation between the sample groups could be observed. As this project continues, we aim to compare the fentanyl-inhibited samples with samples inhibited with compounds of known mechanism. We hope to observe similarities between the fentanyl samples and known mechanism samples that agree with the Lineweaver-Burk results to determine the inhibition mechanism of fentanyl.