Machine learning (ML) is a set of techniques for data analysis used to investigate and predict results from a set of data using algorithms and statistical tools. In pain researches, a lot of data are generated after in vitro and in vivo experiments that, combined, broaden the knowledge about mechanisms and molecular targets for pain control. Behavioral test is one of the most important tools used on pain studies. In animal models, for example, paw withdrawal thresholds and mechanical allodynia are measured with a von Frey equipment before and after a painful stimulus or its treatment. Many of these tests require months of experimentation and involve high-cost drugs and animals. Based on above-mentioned, this work aims to demonstrate how ML techniques could be used to predict how a determined dose of an antagonist would work in a model of neuropathic pain targeting the purinergic signaling. In vivo, experimental test was done using the following antagonist doses: 0.001, 0.01, 0.1, 1.0, and 3.0 mmol. Mean of mechanical allodynia was assessed with electronic von Frey. ML approach was based on Linear Regression and Knn algorithm using scikit-learn implementation; the data set training was configured without 0.01 mmol measures. The resulting ML value was 8.9 g, inside the experimental standard error, and Knn confirmed the data class. ML could successfully predict the mean of animal mechanical threshold. In this context, ML is a very useful tool to guide behavioral experiments, concentrating efforts to predict experimental results and prevent costly tests.

MACHINE LEARNING - FROM ARTIFICIAL INTELLIGENCE TO PAIN RESEARCHES: DELINEATING LABORATORY EXPERIMENTS