Analysis Of Molecular Dynamics Simulation Trajectory Of BamA Protein Based On Machine Learning Algorithms

?-barrel membrane protein is a transmembrane protein,often referred to as ?-barrel outer membrane proteins(OMPs).It was mainly found in the outer membrane of chloroplast and mitochondria of eukaryotes and the outer membrane of gram-negative bacteria.OMPs carries out many biological processes,including transmembrane transport,signal transduction and movement.It is worth noting that it is associated with the release of the virulence,the endotoxin of Gram-negative bacteria.Therefore,it is possible to prevent or treat endotoxin-induced diseases by studying the biogenesis of OMPs.The assembly and insertion of OMPs relies on ?-Barrel Assembly Machinery(Bam)complex.The BamA protein in this study is the core protein of the Bam complex,so studying the BamA protein may be helpful in finding new therapeutic targets.The BamA protein consists of one?-barrel structure and five POTRA(Polypeptide Transport-Associated)domains.Molecular Dynamics simulation is widely used in the research of various field as the crystal structure of proteins is more and more resolved and the computer is rapidly developing.It can make up for the shortcomings of the experiment and provide a new perspective for the study of the problem because it can show the state of protein movement.This method obtains the motion coordinates of the next step by solving the motion equation in Newton's Second Law of Motion-Force and Acceleration,which is essentially a sampling process.The time scale of BamA protein trajectory data in this paper is about 12.5 ?s.Most of the current studies on BamA proteins are based on the relative motion of the ?-barrel structure and the POTRA domains.In this paper,we studied it with the view of residues which is a more detailed aspect.The movement of BamA protein was studied by the perspectives of conformational changes,translation and rotation,global motion of residues and the correlation between conformational changes of residues and conformational changes of whole proteins.Then we found same sites associated with structural transformation of BamA protein.Finally,it maybe provides useful information for clearing the biological function of BamA.Due to the high-dimensional characteristic of trajectory data and its distribution is unknown,we used machine learning algorithm(dimensionality reduction and clustering)to analysis it.Dimension reduction extracts the main features of the data and visualizes the data.Clustering can estimate the distribution of the data and provide ideas for the next step.The dimension reduction algorithm we selected is Principal Component Analysis(PCA),and clustering algorithm is MeanShift and Gaussian Mixture Model(GMM).We calculated the mutual information(MI)among different residues by the clustering results in order to measure the correlation between residues.By looking up the remote correlation between the residues and comparing with the existing studies,we found the A318,T724,W718 and A755 sites and predicted that they are potential binding sites or allosteric sites of BamA which may play an important role in the structural transformation of BamA protein.At the same time,we observed that when the translational rotation and the conformational changes exist simultaneously,the translational rotation is more obvious.