Prediction Of Hot Spots In Proteins Using Digital Signal Processing

Prediction of the locations of hot spots in proteins is an important and difficult area in the field of bioinformatics, it has an important significance for drug design and metabolism analyze. The current method of predicting protein hotspots generally fall into two categories: one is based on experimental method and the other is based on computer technology. In this paper, the application of digital signal processing (DSP) for the prediction of the locations of hot spots in proteins is explored. Since the intrinsic discrete characteristics of the biological sequences, hot spot locations in the protein can be clearly identified by the digital signal processing.Firstly, this paper introduces the insufficient of the method based on single filter, and modifies its deficiency. Proposed one method which based on multi-stage filter, this method designed a multi-stage filter for protein filtered. Identify the locations of the hot spots in protein could through analysis the peak of the energy spectrum of output signal.Again, analyzes the characteristics of protein sequences. In essence, the digital signal of protein sequences obtained by mapping is a non-stationary signal, and time-frequency characteristic of the signal needs to be expressed in local properties in the processing. But the Fourier transform can not express the nature of the time-frequency localization characteristic of the signal. To solve this problem, introduces a new method based on continuous wavelet transform, and designs a fast algorithm for continuous wavelet transform to solve the problem of Fourier transform is not suitable for handling non-stationary signal and the low efficiency of multi-stage filter.Finally, verifies and analyzes the two methods presented in this paper by experiment. Compared the experimental results with existing prediction method and known hot spots in the ASEdb database, shows the correctness and effectiveness of the proposed methods.