Machine Learning

New ML-Based Test Could Improve Detection of Synthetic Cannabinoids

Breaking research published today in AACC’s Clinical Chemistry journal

A novel study published today in AACC’s Clinical Chemistry journal shows that a machine learning-based test detects emerging synthetic cannabinoids with greater ease than standard methods. By making it simpler to identify new synthetic cannabinoids, this innovative screening test could significantly improve the ability of health professionals and lawmakers to protect the public from these potentially dangerous designer drugs.

View the full study here: https://doi.org/10.1093/clinchem/hvac027

With 10-30 new synthetic cannabinoids emerging every year, these substances are one of the fastest growing groups of designer drugs. And unlike the traditional recreational drug cannabis, synthetic cannabinoids tend to be highly active at the CB1 cannabinoid receptor in the brain, which can lead to serious adverse health effects in users and even death. In light of this, there is a strong need for a test that can screen emergency department patient samples for the presence of new synthetic cannabinoids. This could improve treatment for patients presenting with acute toxicity, while also providing public health and legislative officials with the information they need to corral the designer drug market. However, the current gold standard method for identifying these drugs—liquid chromatography high-resolution mass spectrometry (LC-HRMS)—is time consuming and expensive, and therefore not well-suited for widespread use.

To address this problem, a team of researchers led by Christophe P. Stove, PhD, of Ghent University in Belgium, developed a highly accurate screening test for synthetic cannabinoids that is also easy to perform. The test is an activity-based bioassay that uses cells that are specially designed to express the CB1 cannabinoid receptor on their surfaces. When these cells are exposed to patient samples that contain synthetic cannabinoids, the CB1 receptor is activated and the cells emit fluorescent light. A machine learning model then assesses the level of cell luminescence over time to determine which samples are positive for synthetic cannabinoids.

Stove’s team evaluated the efficacy of this method by using it to screen 968 blood samples obtained from adult patients with acute recreational drug toxicity. The results of the screen were interpreted in two ways: 1) by manual expert review and 2) with the machine learning model. The researchers also tested the blood samples with LC-HRMS and then compared the results from LC-HRMS with the results from their method. What they found was that the activity-based screen with manual expert review detected 141 of the 149 samples that LC-HRMS confirmed as positive for synthetic cannabinoids. This means that the activity-based screen has a high sensitivity of 94.6% when the results are interpreted by an expert—a level of sensitivity that the machine learning model was able to match.

“In conclusion, the bioassay continued to demonstrate outstanding performance, confirming its potential as an ideal untargeted screening assay, capable of sensitively and universally detecting new circulating [synthetic cannabinoids],” said Stove. “Although the bioassay itself is already simple and quick, adopting a machine learning approach could potentially speed up sample scoring substantially, reducing the workload, which is ideal for a first-line screening approach complementing conventional analytical methods.”

For more such updates and perspectives around Digital Innovation, IoT, Data Infrastructure, AI & Cybersecurity, go to AI-Techpark.com.

Related posts

ML-driven Medallia Reports Record Q1 Fiscal 2022 Revenue

Business Wire

IARPA Selects BAE Systems to Advance ML to Assure Data Security

Business Wire

integrate.ai Announces Availability of ML & analytics platform

Business Wire