Reducing maladaptive behavior in neuropsychiatric disorders using network modification.

Neuropsychiatric disorders are caused by many factors and produce a wide range of symptomatic maladaptive behaviors in patients. Despite this great variance in causes and resulting behavior, we believe the maladaptive behaviors that characterize neuropsychiatric disorders are most proximally determined by networks of neurons and that this forms a common conceptual link between these disorders. Operating from this premise, it follows that treating neuropsychiatric disorders to reduce maladaptive behavior can be accomplished by modifying the patient's network of neurons. In this proof-of-concept computational psychiatry study, we tested this approach in a simple model organism that is controlled by a neural network and that exhibits aversion-resistant alcohol drinking - a key maladaptive behavior associated with alcohol use disorder. We demonstrated that it was possible to predict personalized network modifications that substantially reduced maladaptive behavior without inducing side effects. Furthermore, we found that it was possible to predict effective treatments with limited knowledge of the model and that information about neural activity during certain types of trials was more helpful in predicting treatment than information about model parameters. We hypothesize that this is a general feature of developing effective treatment strategies for networks of neurons. This computational study lays the groundwork for future studies utilizing more biologically realistic network models in conjunction with in vivo data.