The evaluation of patients with chest pain represents a challenge for health professionals, as heart attack is a major source of deaths in the world. A genetic algorithm (AG) is presented to select the best set of rules that can support its diagnosis. The individuals were represented as a combination of 17 logical OR or AND operations (determined as 1 or 0) that related the 18 variables of the Braunwald scale. A population of 200 individuals was selected and from them 200 children were generated by recombination and mutation (95% and 5% respective probability), during 200 iterations (generations). The correspondence fitness function was calculated from the evaluation of the phenotype of each individual in the training set (119 patients). After validating the rules resulting in the set of tests (40 patients) an accuracy of 85% in the diagnosis was reached. This result is similar to the performance of emergency physicians and could serve as support in the differential diagnosis of acute coronary syndrome. Abbreviations: CE, evolutionary computation; SCA, acute coronary syndrome; AG, genetic algorithm; PG, programming and genetics.