RESUMEN
We have developed classification coefficients and an equation to detect heterozygotes for phenylketonuria. The combination of several variables (Phe, Phe/Tyr, Phe2/Tyr) gave a safe diagnosis in more than 96% of cases. We then computerized a random selection of our population, which was divided into two groups: the first was "selected" to compute discriminant functions, while the second, excluded from computation, was used to check the fitness of our method. Despite the reduction of sample size, 95.2% of unknown subjects were correctly classified. Finally, we used our equation to detect heterozygotes for phenylketonuria in a population of 26 children, affected by non-specific mental retardation, and their mothers. We found a high proportion of carriers for phenylketonuria, defined as subjects having a percent probability of correct classification higher than 90. By this method, heterozygosity was detected in two child-mother couples, four individual children and five mothers.