Abstract: Clustering decoys is a necessary post-processing step for protein structures prediction. There are many clustering algorithms used in protein structures prediction. The Affinity Propagation(AP) clustering algorithm is ingenious as it propagate some information among all the data points, and reach a steady state after some iterations. This paper applies the affinity propagation clustering algorithm in protein structure prediction, and test it in seven different data sets. And then, we compare the exemplars with their native conformations with the RMSD metric. Results show that the Affinity Propagation clustering algorithm is better than Rosetta clustering algotirhm under 67%, and it is fit for protein structure prediction.