Computational Analysis Of Intracellular Protein Stability Based On Sequence-derived Features

Various functional proteins in eukaryotic cell are in a dynamic balance of synthesis and degradation in order to maintain cell normal physiological functions. The ubiquitin-proteasome proteolytic pathway is one of the main ways of intracellular protein degradation. It is a process that the cells modulate its condition which is supposed to influence the cell turnover and metabolism, proliferation and differentiation, signal transduction, and other basic cellular process as well. The target protein identification and degradation in the ubiquitin–proteasome pathway shows the close correlation between protein degradation and features of its sequence and structure. Therefore, we analyzed the influencing factors for the intracellular protein stability basaed on protein sequence, with the purpose of deepening the understanding of the degradation mechanism.After the preprocessing of high-throughput proteomic experiment data, we analyzed the influence factors of proteins'half-lives from several aspects, such as protein physicochemical properties, sequence signature, protein structure, key motifs and protein post-translational modifications. The results indicated that the transmembrane, signal peptide, hydrophobicity and the composition of di-peptide are significantly correlated with the protein stability. Based on the results on various features related to protein degradation, we established a human short-lived protein prediction model named SProtP, which can achieve sensitivity of 80.0% and specificity of 79.3 on test dataset.Some of the key distinctive features are analyzed in detail. By comparison with N-end rule, SprotP demonstrate the effectiveness of our model on predictive ability. We also discussed the SProtP predicting ability by using different amount of features and different threshold of classifying samples respectively in order to obtain simplicity and optimality of SProtP model. Testing on the independent dataset, SProtP shows good stability and generalization ability. In addition, the conservative and evolutionary of protein stability between human and yeast proteome are also discussed in this thesis. The web server for predicting human short-lived proteins is established, adopting the B/S model. The web server interface is programmed by HTML and JavaScript, while network background is composed of CGI scripts programmed in Phython language with the platform of Apache server. According to the web server, the results of this research can be applied wider and more effectively. The URL of this prediction software is free available at: http://reprod.njmu.edu.cn/sprotp/index.py.