Abstract: In DNA computing, one requirement of the encoding DNA sequences is that they should keep similar thermodynamic stability in order to maintain the reliability of the computing process. Melting temperature is a suitable parameter used to evaluate the thermodynamic stability of DNA sequences. As traditional predicting methods used in biological engineering may exist lager error for a few sequences, thus it couldn’t meet situation involved large amount of DNA sequences as DNA computing. In this paper, we introduced a BP Neural Network to predict the melting temperature based on the Nearest-Neighbor parameters. Our result shows that the predicting error can be limited within ±5.5℃.