Based On Structural And Pseudo-Amino Acid Composition Information To Predict The Anti-Apoptosis Proteins And Pro-Apoptosis Proteins

Protein dominates all life activities, and coordinates a series of complex biological processes; however, different types of proteins perform different biological functions. Apoptosis is a protein which has a special function protein, plays an important role in the organism growth and maintains homeostasis. During apoptosis the anti-apoptosis and pro-apoptosis perform different function in the regulation of apoptosis. If the inactivation of anti-apoptosis proteins or pro-apoptosis proteins, it will lead to the occurrence of cancer and other diseases. So the classification of anti-apoptotic proteins and pro-apoptotic proteins will help us understand the pathogenic mechanism of the apoptosis proteins better.In this paper, the apoptosis proteins structure prediction are investigated, based on the construction of new dataset, the search for information of the protein sequences, physicochemical characteristics, and protein structure, we predicted the pro-and anti-apoptosis by using the Increment of Diversity (ID) and Support Vector Machine (SVM) algorithm. The main contributions are summarized as follows:1. We first constructed the pro-apoptosis and anti-apoptosis dataset D361. Using SVM algorithm to predict in Jackknife test the success rate achieved the desired prediction. The jackknife test based on hybrid feature method to predict the success rate is generally higher than other prediction success rate.2. Based on the data set D361, according to the Uniprot website updates, we enriched the data set D461, and we found the positive and negative sets sequence numbers tend to be more balanced than the previous data set. By combine the SVM algorithm, both single feature and hybrid feature to predict the success rate in jackknife test are having greatly improved.3. According to the biological and physicochemical characteristics of apoptosis protein, we extracted several feature information shown as follow:protein sequence information, amino acid hydropathy information, protein block information, Position-specific scoreing matrix (PSSM), chemical shifts information and protein n-terminal sequence component information, and furthermore, the impact of the single feature and multi-feature fusion models on predictive results was analyzed.4. For the first time, the ID and SVM algorithm are applied to the classification prediction of pro-apoptotic and anti-apoptotic proteins, and the predicted results of the two algorithms were compared.