Font Size: a A A

Research On Model Building Algorithm For High-resolution Cryo-Electron Microscopy Maps

Posted on:2019-04-13Degree:DoctorType:Dissertation
Country:ChinaCandidate:N Y ZhouFull Text:PDF
GTID:1360330590951529Subject:Biology
Abstract/Summary:PDF Full Text Request
Cryo-electron microscopy(cryoEM)is one of the most important methods in structural biology.It collects two-dimensional(2D)projections of proteins from vitreous ice using the transmission electron microscopy(TEM).Then dedicated algorithms are used to reconstruct a three-dimensional(3D)electron density map.The atomic model is built from the reconstructed density map at last.Recently,high-resolution electron density maps of small protein with low symmetry can be obtained by cryoEM owing to the development of imaging,algorithms and computing capabilities.However,previous research has shown that voxel size errors of cryoEM density maps can be as much as 5%,which has a strong impact on model building of high-resolution cryoEM maps.Therefore,voxel size of density maps must be refined before model building.The existing method for voxel size refinement requires apriori atomic models,which cannot apply on a density map without any apriori models.The voxel size refinement is an important problem in model building of cryoEM maps.Moreover,the atomic model of high-resolution cryoEM density maps are built manually in current stage.Interpreting cryoEM maps using manual atomic models is a difficult and time-consuming task.No program currently exists for de novo cryoEM model building at high resolutions using an automatic algorithm.In this work,I present statistical templates calculated from cryoEM maps and a series of model building algorithms based on template matching for high-resolution cryoEM maps.These algorithms solve the problem of voxel size refinement for cryoEM maps and build model automatically.Using test results,I found that our templates are good representations of cryoEM map features.Our voxel size refinement algorithm corrected voxel size error from 5% to 0.59% on average and 0.06% in the best case.Furthermore,I found that 0.5% voxel size error had no significant impact on model building.As for automatic model building,our de novo model building algorithms produces better results than other crystallographic model building programs and its elapsed time is much shorter than manual model building.For a 1 MDa protein,the total model building time can be as short as 23.5 hours.In conclusion,this research presents template matching-based model building algorithms for high-resolution cryoEM maps.These algorithms are capable of refining the voxel size errors of cryoEM maps to negligible degree and building the atomic model full automatically.This greatly reduces the workload of manual model building and facilitates the development of high-resolution cryoEM model-building methods.
Keywords/Search Tags:CryoEM, Atomic model building, Voxel size refinement, Template matching
PDF Full Text Request
Related items