Studies On Prediction Of Selective Cleavage Sites And Cleavage Profile Of Proteasome Using VHSE Amino Acid Descriptor

The ubiquitin-proteasome system (UPS) plays a critical role in proteolysis anddegradation in the eukaryotes. It is involved in the degradation of virtually every type ofsurplus, dysfunctional or damaged cellular component, ranging from soluble proteins towhole organelles. Recent studies found that the degradation of UPS is strictly controlledin the time and space, UPS can identify and hydrolyze the regulatory proteins with shorthalf-life, therefore it affects the most of major cellular processes. UPS is also implicatedin a broad array of pathological states, such as cell cycle regulation, DNA transcriptionand translation, tissue differentiation, antigen presentation, tumorigenesis, cancer,neurodegeneration and apoptosis. Recently, due to the roles of proteasome in UPS, thestudies on cleavage site prediction and cleavage profile have arracted considerableinterest in computational biology. These studies are of great significance in exploringthe pathogenesis of related diseases and vaccine design.However, the existing cleavage site prediction methods are mostly based onnonlinear models with little physicochemical meanings. In this paper, VHSE (principalcomponent score vector of hydrophobic, steric, and electronic properties), a novel set ofamino acid descriptors, was used to characterize the source proteins of2650naturalMHC class I ligands obtained from AntiJen database,859T cell epitopes from IEBDdatabase,489digestion products in vitro form IEDB database. First of all, multiplestepwise regression (MSR) was used to delete the redundant descriptors which were notrelevant to the target variable. Then, support vector machine (SVM) was employed toestablish prediction models using linear and RBF kernel functions based on thestructural descriptions of amino acids adjacent to the cleavage site. Following that, wemade in-depth analysis of the cleavage mechanism and selective profiles of proteasomebased on the optimal linear model obtained. This optimal method performed well onpredictive accuracy as well as interpretation. The conclusions derived from this paperare of important theoretical significance and reference values for the designs ofepitope-based vaccines and mechanisms studies of proteasome. The main results are asfollows:①An optimal model SVMMHC-Ibased on MHC-I ligand dataset was obtained, ofwhich the Acc, Sen, Spe, MCC, and the AUC were81.48%,85.80%,76.93%,0.6306and0.8850, respectively. The optimal SVMMHC-Imodel can reflect the common cleavage specificities of the constitutive proteasome and the immunoproteasome. Theresults of SVMMHC-Ishow that the hydrophobic, electronic, and steric properties of theamino acids adjacent to the cleavage site are closely related to the cleavage specificity.The positions P9, P8, P4, P1, P3’, P4’and P5’ have an important effect to cleavage,especially the hydrophobic property of P1, P8and P5’. And we also found thathydrophobic potential difference between upstream and downstream of cleavage sitemay benefit the cleavage process of proteasome. Moreover, the electronic and stericproperties of the amino acids in the P1and P5’ positions can also affect the selection ofcleavage site to some extent.②The optimal model SVMTCEbased on T cell epitope dataset was obtained, ofwhich the Acc, Sen, Spe, MCC, and the AUC were81.77%,86.35%,75.20%,0.6411,and0.8840, respectively. SVMTCEmodel can reflect the cleavage specificities of theimmunoproteasome. The results of SVMTCEshow that the hydrophobic property ofamino acid residue at the P1position has more effects on cleavage specificity of theimmunoproteasome, and its hydrophobic residues can be easily identified by immunesubunit active sites which is helpful to the cleagave of substrates into immune epitopes.The amino acids adjacent to the cleavage site display significant cleavage specificities,and the cleavage specificities of residues at the P9, P8, P3, P4, P2’, P4’ and P5’ positionsmay reflect the interactions between the residues of substrate and the residues adjacentto the active site.③An optimal model SVMVITRObased on in vitro dataset can reflect the cleavagespecificities of the constitutive proteasome. The Acc, Sen, Spe, MCC, and AUC ofSVMVITROmodel were76.99%,84.66%,69.31%,0.5463, and0.8410, respectively. Byanalyzing VHSE weight coefficient of amino acid residues near the enzyme cutting site,we got the following conclusions: Selective cleavage is mainly affected by thehydrophobic, electric and steric properties, especially for the hydrophobic of P1site.From the hydrophobic properties of amino acid residues near the enzyme cutting site, itcan be seen that the optimal cleavage mode of proteasome is as follows: hydrophobicamino is at the upstream of cleavage site and non-hydrophobic amino is at thedownstream of cleavage site. The electrostatic properties of the amino residues localizedat P3, P2and P1sites is beneficial to the cleavage and the electrostatic properties of theamino residues localized at P9, P7and P6sites have negative effects on the proteasomecleavage.④The results of external prediction by the above three models show that the Sen, Spe, and MCC were83.2%,62.5%,0.47(SVMMHC-I);87.1%,49.5%,0.39(SVMTCE);83.7%,61.4%,0.46(SVMVITRO). The prediction performance of three models obtainedin this paper are much better than that of NetChop2.0(Sen=73.6%, Spe=42.4%,MCC=0.16), which is considered as the most optimal predictive method at present.⑤By analyzing the specificity of enzyme cut obtained from our model, we foundthat the specificity of enzyme cut of different types of proteasome has the obviousconsistency: for the amino property, the sequence of the contribution to the selectivity ofcleavage site is hydrophobic, electric property and steric property. The amino residue ofP1site near the cleavage site has the biggest effect to the specificity of enzyme cut. Theamino residue of P5site has weaker effect. The “potential difference of hydrophobicity”between the amino residue at the upstream of cleavage site and at the downstream ofcleavage site is beneficial for the enzyme cut. In addition, the specificity of enzyme cutof constitutive proteasome and immunoproteasome has some differences is due to thedifference of constitute of amino residue near the active site and the ambient. Thedifferences is mainly showed in the such things: immunoproteasome enhance thepossibility of cutting enzyme at the downstream of the hydrophobic amino residue.The steric property of the amino residue would not have strong effect to the enzyme cutof proteasome.