Prediction of hot-spots in protein sequences using statistically optimal null filters

The knowledge of hot-spots locations in protein sequences is crucial for understanding protein functionality. It is known that the hot-spots exhibit a characteristic frequency corresponding to their biological function. In this paper, a new technique using a statistically optimal null filter (SONF) is proposed to predict the locations of hot-spots in proteins. The technique involves detecting the characteristic frequency corresponding to hot-spots of interest. This is achieved using an instantaneous matched filter in SONF which increases the signal-to-noise ratio and the estimation is further improved by using a least squared optimization. Through examples it is shown that the proposed technique is more accurate and reliable as compared to the popular modified Morlet wavelet technique.