Addicted or not? A new machine learning-assisted tool for the diagnosis of addiction-like behavior in individual rats

AbstractBackgroundThe transition from controlled to compulsive drug use occurs in a small proportion of individuals characterizing substance use disorder (SUD). The “3-Criteria” model developed on the operationalization of key DSM diagnostic criteria of SUD has helped to shed light on behavioural and biological factors involved in these divergent trajectories. However, the classification strategy on which the model has hitherto relied puts as much weight on the cohort to which the individual belongs as on their own characteristics, thereby limiting its construct validity with regards to the individual-based diagnostic approach in humans.MethodsLarge datasets resulting from the combination of behavioral data from several of our previous studies on addiction-like behavior for cocaine or alcohol were fed to a variety of machine learning algorithms (each consisting of an unsupervised clustering method combined with a supervised machine learning algorithm) in order to develop a classifier that identifies resilient and vulnerable rats with high precision and reproducibility irrespective of the cohort to which they belong.ResultsA classifier based on K-median or K-mean clustering (for cocaine or alcohol, respectively) followed by Artificial Neural Networks emerged as the best tool reliably and accurately to predict if a single rat is vulnerable or resilient to addiction as operationalized in the 3-Criteria model. Thus, all the rats previously characterized as 0 or 3crit in individual cohorts were correctly labelled as Resilient or Vulnerable, respectively, by this classifier.ConclusionThe present machine learning-based classifier objectively labels single individuals as resilient or vulnerable to develop addiction-like behaviour in multisymptomatic preclinical models of cocaine or alcohol addiction in rats, thereby increasing their heuristic value with regards to the human situation.

Know more about