ABSTRACT
Abstract This paper focuses on developing a reduced distance matrix to improve the computational performance during the protein interactions clustering. This proposed matrix considers as centroids two alpha carbon atoms from a protein structure and stores the distances between these centroids and the other atoms from this same structure. Each row in this matrix represents a database record and each column is a distance value. Through this build matrix, clusters were performed using K-Means Clustering. The precision and performance of this presented technique were compared with aCSM, RID and another distance matrix methodology that considers the distances between all atoms from each protein structure. The results were satisfactory. The reduced distance matrix obtained a high precision and the best computational performance.