Image processing for viruses

1. Freezing of virus grids

• For old Quantifoil: wash with Ethyl Acetate
• Glow-discharge in air for 20 mA, 45 s, negative polarity
• Concentration: 4 mg/ml minimum, 7 mg/ml maximum
• Dilute with buffer 50 mM NaCl, 10 mM HEPES pH 7.3, 0.1 mM CaCl2
• Use manual plunger in cold room
• Apply 4.5 – 5 µl to C-flat or Quantifoil grids
• Blot 3 – 5 s from the front
• Plunge in liquid ethane
• Use within 3 days of freezing

2. EM of viruses

• Use F30, coma-free alignment (can load stored setting)
• C2 aperture: 70 µm
• Spot size: 5 – 7 (depending on age/brightness of gun)
• 300 kV beam, Defocus: 2.0 – 3.5 µm
• Magnification: 59,000
• Both Gatan holders work well
• Wait 0.5 – 1 h after grid insertion
• Time between refills: 3 h, wait 0.5 h after each refill
• Grid useless (contamination) after 10 – 12 h
• Exposure: 1 s (25 – 30 e-/Å2)
• Acceptable drift rate: 5 nm/min (~1 Å/s)
• Detector: film
• Scanner: Zeiss SCAI, 7 µm/pixel, scanned individually (8 bit should be fine)
• polarity: negative to negative, use A. Steward’s program to convert to MRC

3. Image processing of viruses

• Particle picking: native resolution, Signature program (signature.tar.gz)
• 2 x 2 binning for initial orientation search
• CTF with CTFFIND3 and CTFTILT, compare values to verify
• Convert picking coordinates to CTFTILT defocus values with shell (ctftilt.tar.gz), script reads all Signature coordinate files and corresponding CTFTILT log files and appends output to a list with defocus values (3 reals)
• Generate new Frealign parameter file from old file (template) using cut/paste in emacs (this leads to incorrect film number, not used in later refinement)
• Run one cycle of Frealign in Mode = 3, 1 deg search interval, ITMAX = 1, IPMAX = 10, RMIN = 200, RMAX = 20
• Run subsequent cycles of Frealign in Mode = 1, RMIN = 200, RMAX = 20, 15, 10, 8, 6, 5, 4
• Run one cycle of Frealign with FDEF = T, FPART = T
• In all cycles: IEWALD = 0, PBC = 3, FBEAUT = T, FCREF = T, ASYM = I2, keep 90% of best particles
• Filtering & B-factor: resolution cutoff at 0.143, B-factor = -300
• NCS averaging: used RAVE package, see example/script (NCS.tar.gz)
• Use PDB to generate mask (done in RAVE), use Chimera to fit into density and write out new PDB file
• Find transformation matrices (done in RAVE) using B-factor filtered 4 Å map (this was a noisy map at this stage because its resolution according to the FSC was only 6 Å)
• Averaging (done in RAVE)
• Frealign map files could be used without further modification but needed to be recalculated with FCREF = F and subsequent B-factor filtering (-300) with bfactor_cref; for the FSC, averaging was applied to two half maps and these were used with IMAGIC (no additional masking or soft edge) Script (FSC.tar.gz)

4. Asymmetric refinement/reconstruction

• No further refinement, use parameters found in normal refinement of reconstruction
• Select ONE 5-fold axis
• Use program “model.exe” to calculate test projections, apply 2D mask to projections and particle images in location corresponding to region of interest in 3D; then calculate local correlation
• For one 5-fold axis, keep orientation with highest correlation coefficient
• Repeat for remaining eleven 5-folds
• Calculate new 3D reconstruction using each particle image 12 times with the 12 orientations found in previous step: this was done by calculating 12 new reconstructions with the particle in one of the 12 orientations, followed by adding the 12 reconstructions.
• Asymmetry developed after 3 iterations (some images still change orientations after that)
• Scripts: Asymmetry.tar.gz

5. Real-space refinement

• Use NCS-averaged map (MRC)
• Use X-PLOR module on alpha (vienna) compiled by James (he also compiled the newer CNS module on Linux but did not use it), binary (CNS_binary.tar.gz, look in notes for more info on other binaries)
• James’ algorithm is identical to Chapman’s algorithm in 2003 but James added I/O for MRC files
• Start with PDB placed into density using Chimera
• Source ulogin.com in X-PLOR directory to set up env variables
• Used fixed B = 100, let x,y,z coordinates, angles refine freely (with normal geometry restraints)
• ncs.dat: define VP7 trimer from one monomer
• Use target=RRES to use real space refinement
• energy.inp: measure total energy (geometry, correlation), for debugging
• To do refinement:
  
  1. run rigid.inp & emin.inp (rigid body docking) (xplor & emin.log)
  2. run tors.inp (torsion angle dynamics) -> refined structure
  3. run scale.inp to generate model-map correlation, change pixel size to get maximum correlation, might have to repeat tors.inp if significant change in pixel size
• To calculate residue-by-residue correlation coefficient, run script (residue_by_residue.tar.gz)
• For an example with defined rigid units within a molecule, look at example_units.tar.gz