Structure-activity methods applied to NPS amphetamines e cathinones in order to predict risk assessment

New psychoactive substances (NPS) have been created to circumvent legal prohibition of abuse drugs. This phenomenon brings difficulties in detecting the risk of using these new substances since their health effects are unknown. Nevertheless, in silico methods present great potential to study NPS. They are able in make predictions for toxicity in a less time-consuming way, giving a positive feedback since these new drugs are in constant expansion. This study aimed to build a QSAR model able to verify the potential risk of new synthetic recreational drugs derived from amphetamines and cathinones. For this purpose, a data set consisting of 23 derivatives with in vitro affinity against the norepinephrine transporter (NET) receptor was selected. We used geometric molecular descriptors (AM1 level) obtained in free platform ChemDes, and the Ordered Predictors Selection (OPS) for variable selection. A PLS model was obtained with good internal (R2=0.914; SEC=0.167; Q2LOO=0.835; SEV=0.231) and external (average of 7 different test sets: R2pred=0.754; SEP=0.246) statistical features. This model was able to accurately predict the risk range of three previously selected derivatives: methedrone (low), ethcathinone (medium), and methamphetamine (high). These data were used to create a simple free program in JAVA to help those interested in using it to determine the risk of new synthetic drugs of abuse of this class, which can be requested from the authors. These results show the potential of using in silico methods as an important support tool for the forensic sciences.