Study On The Thermostability Of Enzyme Based On Pattern Recognition And Molecular Dynamics Simulation

Protein is the function performer of organism and no life action can go without theparticipation of protein. As the natural biological catalysts, enzymes usually have the highcatalytic efficiency and substrate specificity. However, most enzymes are mesophile whichhave poor thermostability, which greatly restricts the application of the enzymes in industrialproduction. Therefore, to understand the mechanism of protein thermostability and find themethods to improve the protein thermostability have become the hotspots not only in the fieldof computational biology but also in protein engineering. The study of protein thermostabilitycan expand the application scope of enzymes, also can help us better understand therelationship between protein structure and protein function. In this paper, we use patternrecognition and molecular dynamics simulation to study the protein thermostability.Here, we used three different pattern recognition methods to predict the proteinthermostability based on protein sequence or structure features. According the accuracies, wefound the key factors which highly relate to protein thermostability and obtained the bettermethod which can be used to predict the protein thermostability. The prediction accuracies ofK-Nearest Neighbors, Support Vector Machine and Decision Tree C4.5on predictingthermophilic proteins and mesophilic proteins were compared based on sequence features.The result indicated Support Vector Machine is better than the other two methods. The10-foldcross-validation test results of Support Vector Machine showed that the prediction accuracieswere88.8%and88.2%respectively based on amino acid composition and dipeptidecomposition. When we used both amino acid composition and dipeptide composition as theinput features, Support Vector Machine got higher prediction accuracy which is89.3%. ThePrediction accuracies could not only prove that Support Vector Machine is a suitable machineleanring method to predict protein themrostbaility, but also could deduce that amino acidcomposition and dipeptide composition are very important for protein themrostbaility.In order to study the influence of structure features on protein thermostability,secondary structure features, hydrogen bond, salt bridge, accessible surface areas of proteinwere extracted which used to predict the protein thermostability. The10-fold cross-validationprediction accuracies of Support Vector Machine were57.3%,70.5%,77.4%and74.7%.These results indicate that prediction accuracies got by using structure features are not asgood as by sequence features. However, when using salt bridge features, there are71.7%thermophilic proteins and82.8%mesophilic proteins could be correctly predicted which alsomeans salt bridge plays a very important role in protein themrostbaility.In this study, we also use the Molecular Dynamics simulation to study the thermostability mechanism of xylanase A (Sl-XlnA). The conformational dynamics ofxylanase A were studied to identify the thermally sensitive regions. With the increasingstimulation temperature, Sl-XlnA begins to unfold at loop4and this unfolding expands to theloops near the N-terminus. The high flexibility of loop6during the300K simulation is relatedto its function. The intense movements of the310-helices also affect the structural stability.The interaction between α4β5loopand the neighboringα5β6loopplays a crucial rolein stabilizing the region from α4β5looptoα6. The most thermally sensitive region isfrom to loop4. The high mobility of the long loop4easily transfers to theadjacent4and and causes them to fluctuate. And, salt bridges ASP124-ARG79,ASP200-ARG159, ASP231-LYS166formed a "clamp" to stabilize the region which includingα4,β4, β5,β6andβ7.