Using Singular Value Decomposition and Discrete Fourier Transform to Characterize Protein Structure and Build Fast Fold Recognition

In order to extract compact and effective feature to characterize protein structure, this paper presents a feature extraction of protein fold by mapping into 2-D distance matrix which is regarded as gray level image and further analyzed by image processing techniques. Firstly, gray level co-occurrence matrix (CoM) of distance matrix image (DMI) is calculated and its singular values are taken as the first group of features. Next, DMI is transformed into frequency view by discrete Fourier transform (DFT). In succession, the magnitude of DFT coefficients is analyzed by histogram of which seven descriptors are taken as the second group of features. Last, the final feature vector is combined by the two groups of features and further standardized by calculating Z-scores before classification runs. The results compared with other methods show that the presented method can characterize effectively protein structure, and perform efficiently automatic classification of multiple types of folds with the benefit of low dimension, meaningful and compact feature, but also no need of complicated classifier system.