Exploiting Geometrical Properties on Protein Similarity Search

This paper discusses about several combinations of protein similarity measurement-methods, with respect to normalization, spatial partitions, geometrical properties, and distance metrics. We compare the effectiveness of possible combinations to each other. Our experiment shows that the feature based on fractional occupancy outperforms other methods. In addition, merging individual features might also yield good result. A prototype of 3D protein geometrical-similarity retrieval system is built for implementing our approach