The Study On Predictive Algorithm For Nuclear Receptors Protein And Bioluminescent Protein

Nuclear receptors (NRs) are ligand-dependent transcription factor, by binding to the ligand and the regulation of gene expression, which is a kind of spread and specific ligand binds to intracellular signaling proteins. NRs play an important role in the body’s growth, development and a variety of physiological processes such as differentiation and metabolism. Research its function and mechanism of action can help to prevent and control many diseases.Bioluminescent Proteins (BLPs) were living organism, which can emit bioluminescence in different cellular. The principle of light-emitting was used widely in the cell of gene regulation, surveillance and immunity. The biotechnology of bioluminescence protein has been applied in medical.In this paper, we chose two datasets about NRs and BLPs from the published literature, by selecting a variety of feature vector information, three algorithms such as support vector machine(SVM), randomforest(RF) and weighted K-nearest neighbors (WKNN) have been applied to predict the proteins, the prediction efficiency was evaluated by the 5-fold cross-validation. Two layer classifiers were used to predict the type of nuclear receptor proteins, the first layer classifier was to predict whether it s the nuclear receptor protein, the best accuracy was 96.84%. The second layer classifier was proposed to identify it among the eight subfamilies, the better predictive results achieved. Further more, the biological fluorescent protein was predicted by using a variety of feature parameters, when the vector of pseudo-amino acid composition was chosen to predict BLPs with support vector machine, the best accuracy was 74.12%.