Abstract: Protein-protein interactions play a critical role in the living body. In this work, the modified resonant recognition model is applied to predict interactions of the yeast proteins. Signal-to-noise ratio (SNR) is selected as a discriminant parameter. Comparing with the previously used peak value, SNR can show significant differences between the positive data and the random data more clearly without compromising prediction precision. Therefore, the proposed method can solve the over-fitting problem.

Key words: discrete wavelet transform, random data test, signal-to-noise ratio (SNR), yeast proteins, modified resonant recognition model