Understanding and exploiting prediction errors minimization within the brain in pharmacological treatments

Abstract

The human brain can be conceptualized as an inference machine that actively predicts and explains its sensations and perceptions: it makes predictions through a probabilistic model. Such a model is continuously and implicitly updated by the computation and minimization of weighted prediction errors, as shown by numerous studies and experimental results. Nevertheless, such an algorithmic functioning of the brain has not been exploited in the neuropharmacological practice. In this manuscript, we show by theoretical analysis and model fitting of previously published data in two different contexts, how it is possible to increase the effectiveness of neuropharmacological and immunosuppressive drugs, through the modulation of the weighted prediction errors. Moreover, on the basis of the proposed model, we derive an optimized drug administration schedule able to increase the drug effectiveness of one order of magnitude, in psoriasis treatment. We make important testable predictions, evidencing the impact and the potential benefit of prediction errors modulation within the brain, in the pharmacotherapeutic practice. Finally, our results lead to a novel formal theory of implicit learning, and shed lights on the actual roles of classical conditioning and UCS revaluation in behavioral and pharmacological conditioning experiments. The potential practical implications of our results are many: the reduction of drugs side effects; the maximization of the therapeutic outcome; a more effective treatment for chronic pain, certain neuropsychiatric diseases, autoimmune diseases and allergic diseases.

Keywords: Drug effectiveness; Neural gain; Partial reinforcement; Prediction errors; Psychoneuroimmunology; UCS revaluation.