Prediction for signal peptides based on the similarity of global alignment

Based on the similarity of global alignment, is was proposed to predict the signal peptides and then avoid the shortcomings of the prior methods by scaling windows, such as loss of useful information and imbalance problem. The similarity can be proved to be embedded into Euclidean space and support vector machine (SVM) can be applied for classification and prediction. The algorithm was carried out on the popular dataset of Neilsen and also compared with other methods. The results prove the stable and high predictive performance.