Can AI treat opioid addiction?
Artificial intelligence could be used to optimize the design of novel opioid addiction drugs through the mass screening of chemical compound data.
In a recent study, a team of researchers at the Marta Fizirola lab from the Icahn School of Medicine at Mount Sinai Hospital (NY, USA) designed a machine-learning model that can design potential novel drugs to treat opioid addiction by blocking kappa-opioid receptors.
Globally, more than 16 million individuals suffer from opioid use disorder and addiction. Over 120,000 people die from drug overdoses a year, the majority of which involve opioids. In America, drug overdose is a leading cause of accidental death and is considered a worsening epidemic. Despite the extreme scale of the problem, opioid addiction is difficult to treat.
Common opioid drugs include legally prescribed pain relievers such as oxycodone, fentanyl and morphine, and illegal drugs such as heroin. These drugs bind to opioid receptors and trigger the release of dopamine, activating intense reward and pleasure centers, as well as providing pain relief.
Opioid withdrawal is one of the key factors driving opioid addiction. Three major G protein-coupled opioid receptors comprise the opioid system: mu, delta and kappa. When opioids bind to mu-opioid receptors in a brain region termed the locus coeruleus, noradrenaline release is suppressed, leading to low blood pressure, lethargy and slowed breathing. However, with repeated opioid stimulation, neurons in the locus coeruleus become desensitized and compensate by increasing their activity in the absence of opioids. This causes ‘withdrawal’ symptoms due to an excess of noradrenaline, such as anxiety, muscle cramps and insomnia, leading to physical dependence.
Kappa-opioid receptors play a crucial role in the mediation of brain reward pathways and mood regulation, and have been revealed to regulate drug-seeking behavior and relapse in addiction models. Preclinical studies in animal models have also demonstrated that opioid dependence may be treated by obstructing these receptors. Thus, kappa-opioid receptors are an important therapeutic target for opioid addiction treatment.
Unfortunately, the process of drug discovery for targeting the kappa-opioid receptor is slow and inefficient. Screening billions of chemical compounds is tedious and time-consuming.
This is where the team’s novel machine-learning algorithm comes in. By utilizing artificial intelligence to optimize the efficiency of drug screening, the team sifted through vast amounts of information extracted from large chemical databases and discovered patterns to design new drugs from scratch.
The machine learning model was trained using data relating to known drugs and the kappa-opioid receptor, enabling the model to produce potential receptor-blocking compounds. A reinforcement algorithm was applied to reward favorable drug treatment properties.
Until now, the team have discovered numerous promising potential drugs with favorable properties. They hope to assess the ability of the drugs to block kappa-opioid receptors before evaluating their safety and efficacy on animal models, and maybe clinical trials in the future. So far, this technology offers an optimistic pharmacological strategy for opioid dependence treatment.
