Prediction of the Helix/Sheet Content of Proteins from Their Primary Sequences by Neural Network Method

摘要

The amino acid composition and the biased auto-correlation function are considered as features, BP neural network algorithm is used to synthesize these features. The prediction accuracy of this method is verified by using the independent non-homologous protein database. It is shown that the average absolute errors for re-substitution test are 0.070 and 0.068 with the standard deviations 0.049 and 0.047 for the prediction of the content of α-helix and β-sheet respectively. For cross-validation test, the average absolute errors are 0.075 and 0.070 with the standard deviations 0.050 and 0.049 for the prediction of the content of α-helix and β-sheet respectively. Compared with the other methods currently available, the BP neural network method combined with the amino acid composition and the biased auto-correlation function features can effectively improve the prediction accuracy.