Research On The Prediction Of Protein Multi-label Classification Based On Pseudo Amino Acid Composition

Protein is an important material in sustain life activities. In recent years, with the development of biotechnology, the number of proteins shows a great increasing trend. It can't satisfy people's demand for protein structure and function study only depending on the experimental methods. In order to shorten the research cycle and save valuable research funding, biologists are increasing dependecy on computational methods, which can be used to study many biological properties of proteins.As an important branch, the issue about classification of proteins functions and subcellular localizations of proteins has been drew more and more attentions and pursuing of researchers in recent years. However, the current studies all based on the premise that a protein only with one special property (class or type characteristics) or solely exist at a certain organelles. However, with the rapid growth in data volumes, the premise can't reflect the real situation. Because there are many proteins may simultaneously with multiple attributes or present in a variety of organelles. The proteins with these attributes may be play a key role in drug development and treatment of diseases, so these proteins become the important subject of gene drug design and basic research. In order to adapt to the tendency, a new method based on multi-label learning, through integrate the pseudo amino acid composition, which is a common feature extract method, to study the classification of membrane proteins? multi-functional enzyme?antimicrobial peptides and subcellular localization of proteins, and design the predictors, some exciting results were obtained. A higher success rate was obtained in our predictors. I hope this research can play a vital role in promoting the development of bio-medicine, I also hope some research methods can be applied to other study areas.