Previsão de efeitos adversos de medicamentos

Abstract

Nowadays there are lots of people consuming drugs daily, mainly elderly people. All of these drugs were subjected to a set of clinical trials to assess their efficacy, safety and determine associated adverse reactions. However, clinical trials are performed on an extremely small sample when compared to the target population.The adverse reactions associated with a drug may lead to the emergence of new diseases or, in extreme cases, may lead to the death of the patient. Health professionals take into account the adverse reactions publicly available when they want to prescribe a medicine. This led to the idea of a data mining project that intends to manage the information on the active principles of the drugs and the adverse reactions already known, in order to create a model capable of predicting adverse drug reactions.The first approach to the problem makes use of recommended systems, and verifies the similarities between drugs and adverse effects of already known drug-adverse pairs in order to predict adverse effects not yet known by the model.The second approach to the problem takes advantage of predictive algorithms using the molecular descriptors of the drug's active principle in order to find justifications for the appearance of a given adverse drug reaction.Thus, it may be possible to discover new adverse effects without resorting to clinical trials. The latter practice, in addition to putting the lives of the population at risk, is more costly and time-consuming than the proposed alternative.