Ella Newman

Topic

Illicit Opioid Detection Using Vibrational Spectroscopy and Machine Learning

Department of Chemistry

Date & location

• Tuesday, August 18, 2026
• 8:30 A.M.
• Elliott Building, Room 226

Examining Committee

Supervisory Committee

• Dr. Dennis Hore, Department of Chemistry, University of Victoria (Supervisor)
• Dr. Alexandre Brolo, Department of Chemistry, UVic (Member)

External Examiner

• Dr. Kim Venn, Department of Physics and Astronomy, UVic

Chair of Oral Examination

• Dr. Catherine Bachewich, Department of Biochemistry and Microbiology, UVic

Abstract

The unpredictability of the illicit drug supply is a significant contributor to the considerable overdose rates over the past decade. Fentanyl, a synthetic opioid, has been a drug of note due to its high potency and involvement in the majority of overdoses. In addition, illicit opioid samples may also contain fentanyl analogues and other psychoactive adulterants, complicating overdose protocols. Drug checking has been implemented to reduce the harms of the unregulated drug supply by providing people who use drugs with a breakdown of the components in their samples. Fourier transform infrared spectrometry is one of the most common drug checking tools in North America because it is accessible, portable, and capable of performing broad, untargeted library searches. However, the sensitivity of this technique is limited, but due to the subjectivity of library matching, its detection limit in drug checking settings is challenging to define. Detectability also varies across illicit opioid sample components, primarily due to variations in their spectral fingerprints. A detectability index and a method for calculating effective limits of detection were developed to describe the limitations in detecting illicit opioid sample components. Results suggest that analytes in opioid samples are more difficult to detect in caffeine-containing matrices, in contrast to an erythritol-only matrix, due to greater spectral overlap. Detectability indices were consistent with the limit of detection results, supporting their accuracy in practice. Detectability information is beneficial for reporting results, training drug checking technicians, and predicting the ease of detection of new psychoactive substances in the illicit supply.

To highlight the need for trace detection and distinction of fentanyl and fentanyl analogues, concentration and co-occurrence trends in the illicit opioid supply were analyzed. Fentanyl and the two most common fentanyl analogues, para-fluorofentanyl and ortho-methyl fentanyl, were found to be most often present below 5% w/w, and frequently found in combination. To enhance detection ability, binary random forest classifiers were applied to Fourier transform infrared and surface-enhanced Raman scattering spectra. Surface-enhanced Raman scattering outperformed Fourier transform infrared spectrometry and enabled trace detection of fentanyls, both individually and when present within the same sample. However, when fentanyls were present at disproportionate concentrations, competitive binding prevented the detection of fentanyls at lower relative concentrations. This work presents limitations of common drug checking instruments, the relevance of these limitations given the composition of the illicit opioid supply, and evaluates methods for improving the sensitivity of current techniques. Results aim to support community drug checking and the analysis of complex illicit opioid samples using accessible and portable vibrational spectroscopy.