IndexPro: An efficient protein indexing system based on knot theory

Abstract

The proteins, with complex three dimensional structures, are traditionally time consuming to observe and work with. In this paper, we apply knot theory to represent proteins as an easily extensible numerical model, an R30. The model can not only be indexed efficiently with some commonly used spatial access methods in databases, e.g., the R* tree and the M tree, but can also be extended by any feature which can be quantified. We construct the system IndexPro and design several experiments to evaluate its performance. As the experimental results show, our system can correctly classify proteins at a satisfactory success rate. Moreover, with the help of the indexes, our system can operate tens of thousands of times faster than existing systems, and scale quite well as the number of proteins grow.