Evolutionary Analysis And Functional Study Of Nedd4Ubiquitin Ligases Using Sequence Features

Ubiquitin-protein ligases(E3s) play an important role in controlling substrate specificity in theubiquitin-proteasome system(UPS), which regulates many important cell processes in eukaryotes.Nedd4E3s implicated in diverse biological processes including endocytosis, degradation of embraneprotein, control of cell growth and virus budding, and the error regulation of which may lead to manyhuman diseases. Therefore, evolutionary analysis and functional study of Nedd4E3s wereinvestigated at the molecular level in this thesis. The contributions of our work are as follows:We analyzed codon bias of Nedd4E3s, which revealed that codon usage bias of the family was low.According to correlation and correspondence analysis, it was obtained that primary factors of codonbias were content of nucleotides and selection pressure. There were16ptimal codons with almostending of G or C. Further more different species have the similar codon usage patterns since thefamily members perform the same function, while Nedd4E3s with higher bias in yeast than otherspecies mainly contained A-ending and T-ending codons. Then we employed PAML software basedon Bayesian theory to analyze adaptive evolution of Nedd4E3s. The results of existing positiveselection signal only in lower eukaryotes suggest that function of Nedd4E3s has been extendedduring process of evolutionary and stable in higher organisms. In order to improve the accuracy of theresults, we analyzed the functional domains. There were no positive sites detected, indicating thatdomain sequences were conserved during evolution and functions of Nedd4Ubiquitin ligases havebeen well inherited.The approach of recognizing substrates of Rsp5in yeast using sequence features proposed in thispaper provided a valuable alternative for finding E3s, and will help further elucidate the functions ofRsp5. Construction of positive, negative and test datasets was the first step. To investigate themolecular determinants of the substrates, we systematically analyzed many features that might playimportant roles, including recognition motifs of Rsp5substrates, regulation of ubiquitylation, relatingto stability of yeast proteins, classic signals of degradation, and so on. The results of feature selectionshowed that PPxY motif, transmembrane region, disorder region and N-linked glycosylationmodification are the most important features. Then we constructed an SVM-based classifier torecognize the substrates of Rsp5, and obtained a sensitivity of87.7%and a specificity of81.1%,which was significantly better than the method of PY motif. We also employed this model to predictproteome-wide substrates in yeast, of which two substrates are already verified by experiments from recent literatures. Finally, in order to demonstrate the model’s performance, some aspects such as halflife, subcellular localization and function are compared. In the process, to improve efficiency, wedeveloped the integrated network platform of SVC based on B/S model.