Incorporating the depth and protrusion index in the protein ray based descriptor

One of the most popular issues in bioinformatics community nowadays is determination of protein function. In order to achieve that, efficient protein tertiary structure retrieval methods should be provided. The necessity of accurate, robust and efficient algorithms for protein tertiary structure retrieval is even more obvious as the number of known protein structures rapidly increases. In this paper we incorporate several features in the protein ray based descriptor in order to provide more accurate protein tertiary structure retrieval. First, the protein backbone is uniformly interpolated. Then, for each of the interpolation points, beside the Euclidean distance from the interpolation point to the center of mass, the accessible surface area (ASA), relative ASA (RASA), depth index, protrusion index and hydrophobicity of the nearest amino-acid are extracted and incorporated in the descriptor. The experimental results showed that by incorporating these features, higher precision (more than 2% improvement) is achieved than by using the Euclidean distance as a single feature.