Using weighted extreme learning machine combined with scale-invariant feature transform to predict protein-protein interactions from protein evolutionary information

Protein-Protein Interactions (PPIs) play an irreplaceable role in biological activities of organisms. Although many high-throughput methods are used to identify PPIs from different kinds of organisms, they have some shortcomings, such as high cost and time-consuming. To solve the above problems, computational methods are developed to predict PPIs. Thus, in this paper, we present a method to predict PPIs using protein sequences. First, protein sequences are transformed into Position Weight Matrix (PWM), in which Scale-Invariant Feature Transform (SIFT) algorithm is used to extract features. Then Principal Component Analysis (PCA) is applied to reduce the dimension of features. At last, Weighted Extreme Learning Machine (WELM) classifier is employed to predict PPIs and a series of evaluation results are obtained. In our method, since SIFT and WELM are used to extract features and classify respectively, we called the proposed method SIFT-WELM. When applying the proposed method on three well-known PPIs datasets of Y east, Human and Helicobacter.pylori, the average accuracies of our method using five-fold cross validation are obtained as high as 94.83%, 97.60% and 83.64%, respectively. In order to evaluate the proposed approach properly, we compare it with Support Vector Machine (SVM) classifier in different aspects.