Visualizing and Analyzing CryoEM Density Maps

1
Visualizing and Analyzing CryoEM Density Maps

• Matthew Baker
• mbaker_at_bcm.tmc.edu
• National Center For Macromolecular Imaging
• Baylor College of Medicine

2
Electron Cryomicroscopy

• Over 240 cryoEM papers published in 2005 and
  2006
• 330 structures deposited in EBI EMDB since June
  2002
• Single particles, helical assemblies, 2D crystals
• 4-85 Å resolution
• 79 structures better than 12 Å resolution
• 150 structural models in PDB from electron
  microscopy

3
Interpreting CryoEM Density Maps
Isosurface view of a cryoEM density map
Structural model
4
Pathway for Structural Analysis

• Visualize the cryoEM map
• Identify features of interest
• Segment the density map
• Interpret sequence and/or structural data
• Fit models
• Analyze structure
• Build model

5
Visualization

• Simultaneously visualize and manipulate large
  volumetric, geometric and atomic data sets
• Available Tools
• Amira (Mercury Computer systems)
• Chimera (UCSF)
• Iris Explorer (NAG)
• PyMol (PyMol)
• Vision (Scripps)
• VolRover (UT)
• VTK (Kitware)
• VMD (UIUC)
• O (Uppsala)

6
Identifying Features of Interest
2BTV VP3A
6 Å
9 Å
15 Å
lt4 Å
Size Shape
Strands Connectivity
a Helices b sheets
Sidechains
Domains
7
Segmentation

• Accurate extraction of regions of interest
  (automated and manual)
• Available Tools
• AIRS Segment (AIRS)
• Sail (Explorer)
• Amira (Mercury Systems)
• Qsegment (EMAN)
• UTSegment (C. Bajaj)
• Watershed (N. Volkmann)

8
Sequence Analysis

• From the primary sequence of individual proteins,
  identify structural homologues and/or secondary
  structure
• Available Tools
• PFAM
• Blast/Psi-Blast
• PHD
• JPred
• Phyre/3DPSSM
• Swiss Model
• Robetta
• Bioinfo.pl

9
Fitting Coordinate Data

• Quantitative method for localizing atomic models
  (probe) to cryoEM density (target)
• Available Tools
• Foldhunter (AIRS)
• Situs (W. Wriggers)
• COAN (N. Volkmann)
• EMFit (M. Rossman)
• ModEM (Modeller)
• RSRef (M. Chapman)
• NMFF (C. Brooks)
• MMTK (C. Hinson)
• YUP (S. Harvey)

10
Secondary Structure Analysis

• Quantitative method for discovering and
  annotating secondary structure elements at
  subnanometer resolutions
• Available Tools
• SSEHunter (M. Baker)
• Helixhunter (W. Jiang)
• Sheetminer (J. Ma)
• requires subnanometer resolution

11
Hybrid Methods for Protein Model Building

• Build coordinate-based models in the context of
  the cryoEM density map
• Available Tools
• Chimera (UCSF)
• Coot (P. Emsley)
• O (Uppsala)
• Direct modeling requires gt5Å resolution

• Mod-EM (UCSF)
• Rosetta (U. Washington)

12
CryoEM Density as an Evaluation Metric

• CryoEM density can discriminate amongst a gallery
  of models better correlation scores represent
  better models
• Evaluation of models is resolution dependant
• Low resolution provides basic shape
• High resolution provides finer structural details
  (loops, SSE)

Topf, M. , Baker, M.L., John, B., Chiu, W., Sali,
A. (2005) JSB
Baker, M.L., Jiang, W., Wedemeyer, W., Rixon, F.,
Baker, D., Chiu, W. (2006) PLoS Comp Biol
13
Comparative Modeling

• Iterative model building approach evolves
  target-template alignment with low sequence
  identity
• Best models selected based on composite density
  fitting score and statistical potential score
• Best model alignments evolved and new models
  generated
• Optimal for low-sequence similarity templates or
  remotely related proteins

Topf, M., Baker, M.L., Marti Renom, M.A., Chiu,
W., and Sali, A. (2006) JMB
14
ab initio Modeling

• fold a target sequence without a structural
  template using only chemical and physics-based
  constraints

• Single domain or small proteins
• Limiting step is selection of top model from a
  set of decoys

• Generate large number of decoys
• Reduce number of decoys by filtering using cryoEM
  density
• Exhaustively fit and evaluate top decoys in
  cryoEM density
• Assess top decoys based on a composite modeling
  and fitting score

Baker, M.L. et al. (2006) PLoS Comp. Biol.
15
ab initio Modeling Pathway
Data reduction
Model selection
Density map
16
Pathway to Modeling
17
AIRS Structure Analysis Software

• Toolkit for visualization, discovery and analysis
  of cryoEM structure
• 25 Python-based tools
• Graphical interface using UCSFs Chimera
• Online tutorials and help documentation
• Included with EMAN 1.70

http//ncmi.bcm.tmc.edu/software/AIRS
18
AIRS Tools
AIRS Tools

• Data representation
• Pseudoatom
• Skeleton

• Density transformation
• DensityMorph
• Filters
• ModeViewer
• RotateTranslate
• Track Transform

• Structural analysis
• Autocluster
• ElementMatch
• SSEbuilder
• SSEhunter

• Fitting
• Dock Manual
• Foldhunter

• Segmentation
• Segment Markers
• Segment Mask
• Segment Read
• Segment Simple
• Segment Write

• Conversion
• Dejavu2pdb
• Pdb2sse
• Cmm2pdb
• Pdb2cmm
• Pdb2mrc

• Model Building
• Skelpath
• AutoBuild
• PDBBuild

19
AIRS Tools Examples

• Foldhunter
• Map to map fitting
• Coordinate to map fitting
• Articulated fitting
• Multi-peak search
• Smart search
• Filters

• Segment Simple
• Arbitrary segment size
• Spherical and rectangular masks

20
Localizing Secondary Structure Elements

• SSEHunter
• Simultaneous a helix and b sheet detection
• Fast, interactive interface using UCSFs Chimera
• Per-point scoring
• Automatic and interactive element annotation
• VRML and PDB output of annotated elements

M.L. Baker, T. Ju and W. Chiu., (2007) Structure
21
Practical Guide to SSEHunter

• Required parameters
• Å/pix
• Resolution
• Density threshold
• Origin adjustment
• Annotation with SSEBuilder
• Per-psuedoatom selection
• PDB and VRML output

22
SSEHunter Methodology

• Three scoring functions mapped to pseudoatoms for
  visual assessment of score/confidence
• Helix correlation defines helical regions
• Skeletonization defines sheet regions
• Local geometry predicates evaluates local shape
  and curvature

SSEHunter results
23
SSEHunter RDV P8 CryoEM Density
X-ray structure fit to P8 density
Annotated secondary structure
6.8 Å resolution cryoEM density map Z.H. Zhou et
al, 2001, Nature Struct Biol.
24
What is a Skeleton?

• Compact geometric representation of a volume
• Feature preserving
• Sheets are represented as flat surfaces
• Helices and loops are represented as curves
• Topology preserving
• Maintains density connectivity while minimizing
  number of branches and breaks

RDV P8 skeleton
25
From SSEHunter to a Protein Model

• SSEhunter gives spatial arrangement of secondary
  structure elements
• Sequence prediction defines a linear array of
  potential secondary structure elements
• Skeletons define potential paths between
  secondary structure elements