dc.description.abstract | Owing to the abuse of antibiotics, the infection resistance of microbial pathogens against chemical antibiotics increases rapidly. Antimicrobial peptides (AMPs) are essential components of the innate immune system with the lower possibility on the emergence of resistance and produced by virtually all organisms known on earth, hence become the attractive candidates for development as therapeutics. AMPs are able to resist various pathogenic microorganisms, such as viruses, parasites, bacteria, and fungi. However, little research dedicates to differentiate the multiple functional types of AMPs simultaneously or even analyze those features that may highly related to distinguish them. In this study, we construct 8 classifiers under two-stage framework to identify the AMPs with their functional types. Moreover, we adopted forward selection strategy to find some important features that may associate with the functional types of AMPs. In the first stage, the resulting area under curves (AUC) of AMP classifier is 0.9894 on the testing set. In the second stage, the AUCs of parasitic, viral, cancer, mammalian, fungal, gram-positive bacterial and gram-negative bacterial are 0.7474, 0.9397, 0.8150, 0.8515, 0.8533, 0.8725 and 0.8576 on the independent testing set, respectably. Besides, we found that hydrophobicity, normalized van der Waals volume, polarity, polarizability, charge, secondary structure and solvent accessibility in the first residue were important patterns to identify AMPs and non-AMPs. In addition to these seven properties, pseudo amino acid composition was also the important factors to distinguish different functional types of AMPs. We developed a web-server called AMPfun to provide our classifiers for AMP and their functional types prediction. | en_US |