This work presents differential evolution (DE), applied to the problem of protein folding, using the hydrophobic-polar model. Protein folding is a relevant problem in bioinformatics for which many heuristic algorithms have been proposed. DE is a relatively recent evolutionary algorithm, and has been used successfully to several engineering optimization problems, usually with continuous variables. We introduce the concept of genotype-phenotype mapping in DE in order to map a given folding into a real-valued vector. The methodology is detailed and several experiments with benchmarks are done. We compared results with other implementations and DE has shown to be competitive, robust and very promising