Deciphering Protein Structures from Cryo-EM: A Geometric Approach

1
Deciphering Protein Structures from Cryo-EM A
Geometric Approach

• Tao Ju
• Dept. of Computer Science and Engineering
• Washington University in St. Louis

2
Research Highlights
Mean value coordinates
Brain Mapping
Protein Modeling
Dual contouring
SIG 02
SIG 05
Geometric Modeling
Bio-medical Modeling
Topology repair
Mesh repair
Contour Interpolation
Segmentation
SIG 04
SIG 07
3
National Center for Macromolecular Imaging

• Department of Biochemistry and Molecular Biology,
  Baylor College of Medicine, Houston
  (http//ncmi.bcm.tmc.edu)
• Director Dr. Wah Chiu

Computer lab
Wet lab
JEOL 3000SFF electron cryomicroscope
Linux cluster
4
Structural Biology

• Protein a sequence of amino acids
• Folds into a 3D structure in order to interact
  with other molecules
• Protein function derived from its 3D structure
• Identifying protein structure
• Computational
• Imaging-based

5
Protein Imaging

• X-ray crystallography
• Limited to relatively small proteins or proteins
  that are easily crystallized
• Solution nuclear magnetic resonance (NMR)
• Limited to molecules not much larger than 50-60
  kD
• Cryo-electron microscopy (Cryo-EM)
• No need to crystallize proteins
• Suitable for large (megadalton-sized) molecular
  assemblies
• E.g., viruses
• Limited resolution

6
Cryo-Electron Microscopy
Micrographs (2D)
Virus Structure (3D)
Electron Cryo-Microscope
Cryo-EM Imaging
Single-particle reconstruction
7
Computational Problem
Cryo-EM Volume
Protein Structure
8
Computational Problem
Plate
ß-sheet
a-helix
Tube
Cryo-EM Volume
Protein Structure
9
A Geometric Approach
Plate
Surface
ß-sheet
Tube
Curve
a-helix
Cryo-EM Volume
Skeleton
Protein Structure
10
Our Research

• Computing skeletons from Cryo-EM volumes
• GMP 06 SMI 08
• Protein structure identification using skeletons
• Secondary structures
• Structure 07 SPM 07
• Amino acids and atomic model
• Work in progress

11
Computing skeletons

• Iterative thinning of a segmented protein solid
  GMP 06
• More noise-resistant and identifies prominent
  shape features

Segmented Protein
Bertrand et al.
Our result
Protein structure
12
Computing skeletons

• Iterative thinning of a segmented protein solid
  GMP 06
• Drawback shape of the segmented protein depends
  on the segmentation thresholds

Thr0.9
Thr0.7
Raw Cryo-EM Volume
Thr0.5
Thr0.3
13
Computing skeletons

• Segmentation-free skeletonization SMI 08
• Identifies prominent shape features at all gray
  scales

Raw Cryo-EM Volume
Grayscale Skeleton
14
Computing skeletons

• Segmentation-free skeletonization SMI 08
• Identifies prominent shape features at all gray
  scales

MRI Vessel
Grayscale Skeleton
15
Finding Secondary Structures

• SSEhunter Structure 07
• Use skeletal geometry to locate a-helices and
  ß-sheets

16
Finding Secondary Structures

• SSEhunter Structure 07
• Use skeletal geometry to locate a-helices and
  ß-sheets

Scored Psuedo-atoms
Located SSE elements
Skeleton Score
Helix Score
17
Finding Secondary Structures

• SSEhunter Structure 07
• Use skeletal geometry to locate a-helices and
  ß-sheets

Scored Psuedo-atoms
Located SSE elements
Skeleton Score
Helix Score
18
Finding Secondary Structures

• SSEhunter Structure 07
• Use skeletal geometry to locate a-helices and
  ß-sheets

Scored Psuedo-atoms
Located SSE elements
Skeleton Score
Helix Score
19
Finding Secondary Structures

• SSEhunter Structure 07
• Sheet accuracy
• 100 for sheets with 3 or more strands
• 29.2 for sheets with 1 or 2 strands
• Helix accuracy
• 99.3 for helices with 9 or more amino acids
• 56.1 for helices with fewer amino acids

20
Matching Secondary Structures

• Matching helices in the protein sequence with
  those found in the Cryo-EM volume
• Graph matching between skeleton and sequence SPM
  07
• 4 secs for matching 20 helices Wu 05 takes 16
  hours for matching 8 helices

Cryo-EM Skeleton and helices found by SSEhunter
Helices matching and backbone
Protein Sequence
21
Where we are now
We are here
Would like to get here
Can we get here?
Cryo-EM
SSEs
Matched Helices
Matched SSEs
Atomic Model
22
On our way

• Building atomic models (future work)
• Template-based (comparative) modeling
• Utilize skeletons for searching structural
  homologue
• Skeleton-based flexible deformation of homologues
  onto cryo-EM
• Template-free (ab initio) modeling
• Constraining protein folding by the skeleton
• Interactive modeling
• Placing amino acids along skeleton

23
Open Problems in cryo-EM
HSV-1

• Segmentation
• From a large molecular assembly to individual
  protein domains
• Protein dynamics
• Deformation between two cryo-EM of the same
  protein at different states

Phage
Procapsid
24
Acknowledgement

• Collaborators
• Baylor College of Medicine (Houston)
• Dr. Wah Chiu, Dr. Matthew Baker
• University of California (San Francisco)
• Dr. Andrej Sali
• Washington University (Seattle)
• Dr. David Baker
• Rice University (Houston)
• Dr. Lydia Kavraki
• Funding
• NSF grant IIS-0705538