Prediction of pathogenic and antigenic proteins and peptides by machine learning approach

Abstract

<P>SUCCESSFUL DESIGN OF A VACCINE IS STRONGLY DEPENDENT ON FIRST KNOWING WHAT KIND OF IMMUNE RESPONSE IS PROTECTIVE AND TO WHICH ANTIGEN THIS PROTECTIVE RESPONSE IS DIRECTED. THIS FACT MAKES THE ANTIGEN DISCOVERY A PREREQUISITE AND DOMINANT STEP IN DEVELOPMENT OF IMMUNOLOGICAL RESEARCH. BESIDES, KNOWLEDGE OF PROTEIN FUNCTION, BIOMOLECULAR INTERACTION AND OTHER BIOLOGICAL PROCESSES ARE ALSO ESSENTIAL FOR THE STUDY OF PATHOGENIC MICROORGANISMS SUCH AS VIRUSES AND BACTERIA. IN ORDER TO FACILITATE THE WET-LAB EXPERIMENTS FOR IDENTIFICATION OF ANTIGENIC SOURCES, IN THIS STUDY, WE EXPLORE SEVERAL MACHINE LEARNING METHODS FOR PREDICTING FUNCTIONAL CLASSES OF PROTEINS AND PEPTIDES FROM A VARIETY OF SEQUENCE-DERIVED STRUCTURAL AND PHYSICOCHEMICAL PROPERTIES INDEPENDENT OF SEQUENCE SIMILARITY. OUR RESULTS SUGGEST THAT MACHINE LEARNING METHOD IS A PROMISING USEFUL TOOL FOR IDENTIFICATION OF ALLERGENIC PROTEINS AND ANTIGENIC PEPTIDES, WHICH MAY PROVIDE IMPORTANT INFORMATION FOR ALLERGIC PREVENTION