Skip to main content

Verified by Psychology Today

AI Predicts New Psychoactive Substances

New AI deep generative model identifies structures of future designer drugs.

GJD/Pixabay
Source: GJD/Pixabay

Identifying “designer drugs,” or so-called “legal highs” quickly is a public health challenge plaguing hospital emergency rooms, forensic labs, and law enforcement agencies globally. These are synthetic substances made in a laboratory by nefarious chemists in order to simulate the effects of highly addictive substances such as opiates, ketamine, hallucinogens (psychedelics), cannabinoids, and cathinones. To address this public health problem, researchers at the University of British Columbia published a new study showing how an artificial intelligence (AI) deep neural network called DarkNPS is able to predict future designer drugs before they emerge, a technology that may help save lives.

“Over the past decade, the illicit drug market has been reshaped by the proliferation of clandestinely produced designer drugs,” the study authors wrote. “These agents, referred to as new psychoactive substances (NPSs), are designed to mimic the physiological actions of better-known drugs of abuse while skirting drug control laws.”

According to the American Addiction Centers, serious complications that may arise from synthetic cannabinoid use include psychiatric effects such as suicidal thoughts and thought disorder that may last for weeks or months. Moreover, synthetic cathinones can not only cause dependence and addiction, but also may be more dangerous than the illicit substances (MDMA, cocaine, methamphetamines) they are designed to imitate.

Currently, identifying seized substances is a lengthy process for law enforcement, forensics, customs laboratories, and toxicologists. In efforts to address this, the University of British Columbia researchers sought to use AI to “imagine” and create a database of entirely new psychoactive substances in order to speed up and automate the identification process of future unknown designer drugs.

The researchers trained a deep neural network called DarkNPS on the HighResNPS dataset of over 1,700 unique known chemical structures of new psychoactive substances. HighResNPS is a crowd-sourced database for screening new psychoactive substances based on high resolution mass spectrometry. According to the researchers, it is considered the most comprehensive and current dataset of NPS structures; contributions come globally from forensic laboratories when new substances are discovered, and from law enforcement seizures or biological samples. DarkNPS generated 8.9 million unique molecules.

Based on established metrics, the researchers decided to use a LSTM (long short-term memory network) which slightly outperformed GRU (gated recurrent unit) to use for their recurrent neural network-based architecture. They validated DarkNPS with over 190 NPS and found the algorithm predicted over 90 percent of new psychoactive substances that later appeared on the illicit market after the model was trained.

According to a recent statement by the University of British Columbia, the AI algorithm is currently being used by the United Nations Office of Drugs and Crime, the European Monitoring Centre of Drugs and Drug Addiction, the Federal Criminal Police Office of Germany, and the U.S. Drug Enforcement Agency.

Copyright © 2021 Cami Rosso All rights reserved.

advertisement