Antimicrobial peptides are the endogenous short peptides, which synthesized from the host cells, while defend to the microbes. Those are the important targets to the drug industries, which aid to produce the multifunctional drugs for various infectious disease. Another side, the machine learning methods revolutionized the data-science field, which facilitated more automated prediction from the various problems, which also highly contributed to AMP prediction problem widely. So far, the methods i.e., Support vector machine (SVM), artificial neural network (ANN), Random forest (RF), and discriminant analysis are widely used with various features. In AMPEP proposed method, author identified minimal features, i.e. 23 features, which are more effective for RF based machine learning AMP prediction. The AMPEP method outperforms the state-of-art methods i.e., iAMPpred, and iAMP-2L. In this study they have conducted the test, that which of feature are more effective for the prediction accuracy, i.e., Composition descriptor set (C), Transition descriptor set (T) and Distribution descriptor set (D) and they found D is more effective than other. Within D, the following amino acid properties are highly supportive to the AMP and the NON-AMP classification problem.
Highly contributed amino acid properties #
1) Hydrophobicity 2) Normalized van der Waals volume 3) Polarity 4) polarizability 5) Charge 6) Secondary structure 7) Solvent accessibility
- Prepare muti-fasta file
- The MATLAB script for AMPEP which can be downloaded from http://cbbio.cis.umac.mo/software/AmPEP/
Bhadra, P.; Yan, J.; Li, J.; Fong, S.; Siu, S.W.I. AmPEP: Sequence-based prediction of antimicrobial peptides using distribution patterns of amino acid properties and random forest. Scientific Reports 2018, 8, 1697, doi:10.1038/s41598-018-19752-w.