Due to the large number of protein structures whose functions are unknown, it becomes increasing important to study the structural characteristics of catalytic residues. Here, we use a novel method to calculate the local structural rigidity (LSR) of protein. Based on a dataset of 760 proteins, the results show that catalytic residues have distinct structural properties. They are shown to be extremely rigid based on the calculation of LSR. Finally, we present a machine-learning based method to predict catalytic residues from protein structure using LSR as primary input feature. The prediction sensitivity and specificity are 0.86 and 0.86, respectively, and the Matthew¡¦s correlation coefficient is 0.72.
展开▼
