AI identifies behavioral traits that predict alcohol preference during adolescence

A new study utilizing artificial intelligence has identified specific behavioral patterns that predict alcohol preference in adolescent mice. The findings indicate that sensitivity to natural rewards and lower levels of sociability are strong indicators of alcohol consumption during this developmental stage. In contrast, these behavioral traits did not predict alcohol preference in adult mice. These results were published in the journal Alcohol: Clinical and Experimental Research.

Adolescence represents a distinct period of brain development characterized by significant changes in neural structure. This phase often coincides with the initiation of alcohol use, which can lead to long-term health issues and dependency. Clinical observations suggest that teenagers display a complex mix of behaviors that may increase their likelihood of experimenting with drugs. Traits such as risk-taking, anxiety, and how one responds to rewards are often linked to substance use.

Previous animal research has produced mixed results regarding which specific behaviors consistently lead to higher alcohol intake. Some studies link high anxiety to drinking, while others find no connection or even an inverse relationship. These inconsistencies may stem from the fact that earlier studies often analyzed single behaviors in isolation. Real-world susceptibility likely involves a combination of multiple traits interacting with one another.

The researchers sought to clarify these inconsistencies by looking at multiple behaviors simultaneously. They utilized a machine learning algorithm to analyze a combination of traits to forecast which mice would prefer alcohol. This approach allows for a comprehensive view of how different personality aspects interact to create vulnerability. By comparing adolescents to adults, the team also intended to see if the predictors of alcohol use change as the brain matures.

The research team employed two different strains of mice to ensure the results applied to both genetically similar and genetically diverse populations. The sample included C57BL/6 inbred mice and Swiss outbred mice. The animals were divided into two age groups: adolescents starting at postnatal day 40 and adults starting at postnatal day 120. The adolescent group consisted of 46 mice, while the adult group included 79 mice.

Over a period of three days, the mice underwent a battery of behavioral tests to establish their phenotypic profiles. The researchers first assessed novelty-seeking behavior using a hole-board test. In this assessment, the animal was placed in an arena with holes in the floor, and the researchers counted how often the mouse dipped its head into the holes.

Anxiety levels were assessed using an elevated plus maze. This apparatus consists of two open arms and two enclosed arms raised above the floor. The researchers measured the percentage of time the mice spent in the open arms, which indicates lower anxiety, versus the closed arms.

Social behavior was evaluated with a three-chamber sociability test. This test measured how much time the mice spent in a chamber containing an unfamiliar mouse compared to an empty chamber. This metric provided a clear picture of the animal’s natural inclination toward social interaction.

Coping behavior was tested using a forced swimming test. The researchers observed the animals in water and measured the time spent on active climbing behaviors. This test is commonly used to evaluate how an animal responds to inescapable stress.

Finally, the researchers measured the animals’ response to natural rewards. The mice were housed individually and given a free choice between a bottle of water and a bottle containing a sucrose solution. The amount of sugar water consumed relative to plain water served as a measure of the animal’s sensitivity to pleasure and natural rewards.

Following these behavioral assessments, the mice entered a five-day alcohol preference phase. They were housed individually and given free access to two bottles. One bottle contained water, and the other contained a 10 percent ethanol solution. The researchers calculated the preference for ethanol by comparing the amount of alcohol consumed to total fluid intake.

The team then used a machine learning technique known as pattern regression. They split the data into training sets and testing sets to teach the computer model. The goal was to see if the model could learn relationships between the behavioral profiles and subsequent alcohol consumption well enough to predict the drinking habits of mice it had not yet analyzed.

The machine learning model successfully predicted alcohol preference in the adolescent mice based on their behavioral traits. The correlation between the predicted preference and the actual preference was statistically significant. However, the model failed to find any predictive patterns for the adult mice.

Among adolescents, two specific behaviors stood out as the strongest predictors. The first was sucrose preference, which had a positive predictive value. Mice that consumed more sugar water were significantly more likely to consume higher amounts of alcohol later in the experiment. This suggests that a heightened sensitivity to natural rewards translates to a higher drive for alcohol.

The second key predictor was sociability. The analysis revealed a negative relationship between social behavior and alcohol intake. Adolescent mice that spent less time interacting with other mice were more likely to show a preference for alcohol. This implies that lower sociability acts as a risk factor for increased drinking in this age group.

Other factors included in the model did not contribute as heavily to the prediction. Anxiety levels and novelty-seeking behavior had a lower impact on the model’s ability to forecast alcohol preference. This contradicts some previous theories that suggest anxiety is a primary driver of adolescent substance use.

The study provides evidence that the drivers of alcohol consumption may differ fundamentally between adolescents and adults. The failure of the model to predict adult drinking suggests that behavioral traits established in adulthood do not dictate alcohol preference in the same way they do during development. This implies that the adolescent brain is in a unique state of vulnerability.

The strong link between sugar preference and alcohol intake suggests that the brain’s reward system plays a central role during adolescence. This points to potential involvement of the dopamine system, which processes reinforcing stimuli. The authors also suggest that the orexin system, which regulates reward-seeking and feeding, could be a relevant biological mechanism.

The finding regarding low sociability highlights the protective nature of social interaction. It raises the possibility that oxytocin, a hormone involved in social bonding, might influence alcohol reward during this developmental stage. Adolescents may be more sensitive to the social effects of alcohol, or social isolation may drive a compensatory need for the pharmacological effects of ethanol.

There are several limitations to consider when interpreting these findings. The study utilized a relatively small sample size, particularly in the adolescent group. While machine learning is a powerful tool, it typically requires large datasets to be most robust. The researchers mitigated this by using cross-validation techniques, but larger studies are necessary for confirmation.

Additionally, the study was conducted on mice, meaning the findings do not directly translate to human behavior. The environmental conditions for the mice were highly controlled, which differs from the complex social environments human adolescents experience. Humans also face cultural and peer pressures that cannot be replicated in a rodent model.

The specific strains of mice used may also influence the results. While the researchers used both inbred and outbred strains to increase generalizability, genetic factors still play a role. Different strains often exhibit different baseline levels of anxiety and sociability.

Future research is needed to verify these results with larger groups of animals and potentially different species. Scientists may also investigate whether targeting the orexin or oxytocin systems could help reduce alcohol vulnerability in adolescents. Understanding the biological basis of these behavioral predictors could eventually lead to better prevention strategies for teens.

The study, “Behavioral profile predicts ethanol preference in adolescent mice, but not in adults: A machine learning approach,” was authored by Liana C. L. Portugal, Bruno da Silva Gonçalves, Emily de Assis Fagundes, Maria Fernandes Freire de Sá, Cláudio Carneiro Filgueiras, Ana Carolina Dutra-Tavares, Alex C. Manhães, Yael Abreu-Villaça, and Anderson Ribeiro-Carvalho.