Abstract
For all the information needed to create thermal tolerance is encoded in the protein sequence as proteins of both thermalphiles and mesophiles are composed of the same 20 amino acids. Predicting sequence-depend changes in protein stability is one of the greatest challenges in molecular biology and bioinformatics. The number of protein sequences is exploding as a result of genome and other sequencing projects, and the increasing number of protein is much greater than the increasing number of known protein structures. In this work, we use classical statistical method to determine the correlation between Tm value of protein and its dipeptide. After statistical calculation, we form out a reference table to predict Tm value of protein, and built a web-based interface program on our web site. The accuracy of prediction of our method is up to more than 97.5% in all groups including material data set. For another evidences to confirm the accuracy of our Tm table, we also apply our Tm table to predict about 48 hyperthermal and common genomes, and got a clear boundary of predict results.