After cryo_fit2, secondary structures of my input molecule tend to be broken.

• Doo Nam observed that when he provided a little less ideal helix (i.g. one H-bond length within helix is rather long), then cryo_fit2 tends to "loosen/unwind" that "kink" area.
• When he provided a longer distance_cut_n_o (i.g. 4~4.5), the geometry of helix was kept well (it maintained the helical kink spot).

• For example, phenix.cryo_fit2 <user.pdb> <user.map> resolution=<reso> secondary_structure.protein.distance_cut_n_o=4.5

• Alternatively, a user can specify a lower map_weight_multiply.

• For example, phenix.cryo_fit2 <user.pdb> <user.map> resolution=<x> map_weight_multiply=0.2.

• Alternatively, a user can use phenix.model_idealization after cryo_fit2 to restore broken secondary structures (if any).

How can I provide gaussian filtered maps?

• Many people in flexible fitting field know that often providing gaussian filtered maps results in better fitting.

• For example, flexibly fit with a 4 angstrom filtered map.
• Then with the resultant atomic model and a 2 angstrom filtered map, do flexible fitting.
• Then with the resultant atomic model and a non-filtered map, do final flexible fitting.

• UCSF Chimera can generate these filtered maps.

• Tools -> Volume Data -> Volume Viewer -> Tools -> Volume Filter -> (Set Width, e.g. 4,2) -> Filter
• File -> Save map as -> (name a filtered map) -> Save

How to improve initial cc?

• phenix.dock_in_map, UCSF Chimera's 'Fit in Map (Tools -> Volume Data -> Fit in Map)', UCSF Chimera's manual fitting (using a mouse) often improve initial cross-correlation (cc) between map and model. UCSF ChimeraX's isolde may improve initial cc as well.

How to maximize cc?

• Generally, cryo_fit2 which is followed by phenix.real_space_refine tends to maximize cc.

• If there is a room/chance to idealize secondary structure further even after cryo_fit2, then cryo_fit2 + phenix.model_idealization + phenix.real_space_refine combination resulted in the maximum cc.

• When the fitting needs to pass through a local minimum energetically for a perferct fit (half of the molecule is already fitted, the other half is not fitted), running phenix.dock_in_map or UCSF Chimera's 'fit in map' before cryo_fit2 is recommended.

• When fitting protein molecule to a map requires a significant conformational change,

  • enforcing stronger map_weight (higher map_weight_multiply) tends to fit better (obviously).

  • For example, phenix.cryo_fit2 <user.pdb> <user.map> resolution=<x> map_weight_multiply=15
  • At this high map_weight_multiply, the secondary structures may be broken (slightly or seriously) right after cryo_fit2 running.
  • However, phenix.model_idealization which is followed by phenix.real_space_refine perfectly restored ideal secondary structures and fitted very well.

  • Alternatively, HE_top_out=True sometimes (not always) boosts cc well even with moderately high map weight (confirmed with 6 benchmark cases).

  • For example, phenix.cryo_fit2 <user.pdb> <user.map> resolution=<x> map_weight_multiply=10 HE_top_out=True
  • HE_top_out=True uses top_out potential (rather than harmonic potential) to helix (H) and sheet (E) for more flexible fitting.

I have a SAXS based cryo-EM map.

• Please use map_weight < 0.2 for SAXS (small angle x-ray scattering) data based cryo-EM map
  • A user can transform saxs data into cryo-EM map (mrc/sit) by Situs SAXS
  • However, this is nothing but central points by SAXS, although UCSF Chimera will visualize this "cryo-EM" map (mrc/sit) as regular cryo-EM map

I have a very small molecule to fit.

• When Doo Nam tried to fit a very small molecule (i.e. 11 residues long), both real_space_refine and cryo_fit2 couldn't fit to cryo-EM map.
• The reason could be that effect of sidechains is too large (in this kind of small molecule), so that global fitting is hindered.
• Doo Nam confirmed that phenix.dock_in_map fits globally much better for this case.

Is there a command line option for cryo_fit2 processing on multiple cores?

• (Please be advised that using multiple cores for cryo_fit2 is only useful when explore=True)
• Like other PHENIX applications, specify nproc.

• For example, phenix.cryo_fit2 nproc=40 model.pdb model.map resolution=3

• If nproc is not specified, cryo_fit2 will use 'available number of cores'/3

My input (before cryo_fit2) and output (after cryo_fit2) pdb files have different molecule sizes

Although pdb text files show similar SCALES, input (before cryo_fit2) and output (after cryo_fit2) pdb files may show different molecule sizes in Pymol when Doo Nam used map that was made from phenix.map_box

Troubleshooting is ongoing.

In the meantime, please use UCSF Chimera to visualize bio-molecules instead.

My model is shifted after phenix.model_idealization.

• To prevent this, it is recommended to add map as well.

• For example, phenix.model_idealization <user.pdb> <user.map that was used for cryo_fit2> (resolution information is not required)

"RuntimeError: cctbx Error: Miller index not in structure factor map"

• When a user sees above message, consider to enter a correct resolution.

• For example, Doo Nam uses either EMDB reported resolution or phenix.mtriage derived resolution to properly run cryo_fit2.

• When a user is not sure about correct resolution, often using a lower resolution works.

• For example, when Doo Nam saw an error with resolution=3, then using resolution=5 often solved the problem.

"Sorry: Crystal symmetry mismatch between different files."

• Sometimes, Doo Nam observed,

• "Sorry: Crystal symmetry mismatch between different files.

  (378, 378, 378, 90, 90, 90) P 1

  (103.95, 91.35, 89.25, 90, 90, 90) P 1"

• The first line (378, ...) shows unit cell parameters from model.

• The second line (103.95, ...) shows unit cell parameters from map.

• Solutions

1. Just erase CRYST1 header information from an input pdb file. Then, provide that input file into cryo_fit2.

• Cryo_fit2 will automatically extract CRYST1 header from map (both from original map and phenix.map_box derived map), and prepend (write at the first line) to the pdb file.

2. (Alternatively) Just leave correct CRYST1 header only in an input pdb file.

• For example, when a wrong CRYST1 header exists above a correct CRYST1 header, above error message appeared.

When cryo_fit2 is useful?

• To model dynamic conformational change, cryo_fit2 is useful.
• To simply fit to cryo-EM map, cryo_fit seems to be useful when initial cc_mask and cc_box < 0.6. When initial cc_mask and cc_box >= 0.6, phenix.real_space_refine tends to do similar fitting with faster speed.

What to do after cryo_fit2?

• If there is a room/chance to idealize secondary structure further even after cryo_fit2, then run phenix.model_idealization which is followed by phenix.real_space_refine.
• Otherwise, just run phenix.real_space_refine to improve overall geometry (such as rotamers).