Zoom Transcript

---

Do generally the models have lower resolution than the cryo-EM-maps? If true, then Why?

---

https://www.youtube.com/c/phenixtutorials

---

Phenix citation:

---

https://journals.iucr.org/d/issues/2019/10/00/di5033/

---

Phenix online documentation:

---

https://www.phenix-online.org/documentation/

---

Will cryo-EM replace X-ray crystallography in the near future? Should I begin to study/do cryo-EM? What do you guys think?

---

not replace, join. and yes, you should begin doing it

---

I personally don’t think so. The techniques are complementary. I do think that they may be used in more focused ways - crystallography is great for looking at chemical detail at higher resolution (and in high throughput). Cryo-EM will be used a lot for membrane proteins, larger complexes, and even modest sized molecules.

---

As IIona said, you (everyone) should be doing cryo-EM as well as crystallography!

---

Does sharpening change the resolution of the map?

---

Tom’s reference on the sharpening: https://onlinelibrary.wiley.com/iucr/doi/10.1107/S2059798318004655

---

Set up tutorial data ==================== New project/Set up tutorial data/Select a dataset Choose: Major capsid protein of group A rotavirus (Autosharpen map) Hit OK once or twice to set up tutorial On left panel of Phenix GUI hit "last modified" once or twice to show the tutorial with a check mark next to it.

---

Note that many of the tutorials have a README file you can open when you set up the tutorial, and that gives you some hints about what you’re trying to accomplish and how to do it!

---

Viewing of directories is not supported for this desktop environment (Fedora 32)

---

Sharpening will act as a ‘’filter’’ that will give certain weights to the coefficients, boosting some of them and down weighting others, trying to find a compromise between this ‘’too much detail’’ that Tom was mentioning and amplifying the noise or having a too blurred map difficult to build a model in.

---

Examine starting map ==================== Select Coot (top of page)/File/Open coordinates/rotavirus.pdb then: File/Open map/rotavirus_blurred_map.ccp4

---

what map formats does phenix open?

---

Sharpening doesn’t change the resolution of the map - changes the interpretability as Claudia says. Tom will talk about another method that does change the resolution.

---

Phenix opens MRC, CCP4 format maps.

---

@Ilona: Major formats such as .mrc and .ccp4 are supported

---

so currently, the way to open a map with a model is on coot - I am assuming you are working on doing it with chimera?

---

Run auto-sharpening ======================== In GUI: Select Cryo-EM/Autosharpen map... Non-default parameters: Title: Rotavirus auto-sharpen 2.6 A Map file: rotavirus_blurred_map.ccp4 Resolution: 2.6 Run When done...in your Coot window: File/Open map/Autosharpen_1/sharpened_map.ccp4 Compare auto-sharpened map with original.

---

DNS6 and X-plor/CNS maps are supported too

---

Yes, we are working on connecting the Phenix GUI with ChimeraX. That let’s the Phenix GUI open models and maps automatically.

---

will it sharpen without a model? or is it influenced by the model?

---

The model file is optional.

---

Model will not be required

---

You can use a model if you have one, but it it usually done without a model.

---

Is the resolution you specify under referring to the global resolution or maximum local resolution?

---

I think it is the best resolution (minimal nominal one from the map)

---

@Ilona But currently, you would have to manually open the output map and model in ChimeraX.

---

Using the gold standard FSC resolution is reasonable.

---

How does local sharpening work?

---

For resolution cut off should we use Nyquist limit or FSC at 0.143?

---

if I understand correctly, relion for example already produces a b-factor sharpened map. in that case does it need to be sharpened again?

---

Does the resolution value entered here affect the sharpening? What if the value entered here was 2A or 3A?

---

I guess it will affect as Tom mentioned it is used as a limit

---

It is okay to try - sharpening doesn’t remove or add information - it should be a reversible process (or close to).

---

FSC at 0.143 seems to be the common current practice, but there is still debate I know.

---

Is there any reason to convert a cryo-EM map to structure factors and open as an .mtz file?

---

Which B-factor should be mentioned after sharpening when we report in the publication?

---

If you are using half-map-based local sharpening, should you input the ‘original’ half-maps (i.e. unsharpened and unfiltered) or filtered half-maps? Does the same apply for local sharpening of full maps (when not half-map-based)?

---

I’m also curious about the sharpening as a ‘’reversible’’ process…. In this case because it is phenix algorithm yes but what if other software’s method to apply the factor correction is considering other aspects? I’m also asking because there has been a lot of discussion about making ‘’mandatory’’ to deposit unsharpened maps precisely because ‘’unsharpening’’ is not that easy

---

In Phenix we don’t advocate conversion of maps to structure factors as this is a process that loses information.

---

“Is there any reason to convert a cryo-EM map to structure factors and open as an .mtz file?” — No, unless a particular problem cannot read a map directly

---

How does auto sharpening in phenix compare to post processing in relion?

---

If we use a model for auto map sharpening in this case, do we need to consider symmetry of the model? In this case, do we need to have full virus model?

---

do we need to choose initial b_iso value? And how do we decide?

---

Is there any equivalent tool of Autosharpen for crystallography?

---

@Jaime: No need to set initial b_iso

---

Can we also use the half maps for sharpening?

---

Does the sharpening improve the resolution number we report?

---

In theory sharpening which just applies a B-factor (and a scale) is reversible. However, many of the sharpening algorithms are more complex - Tom mentioned the high-res cutoff, Relion using information from the reconstruction process, these make it more challenging to reverse.

---

Sharpening doesn’t change the resolution you report.

---

I see, thanks Paul.

---

so would you take a refinement job output for autosharpening?

---

is any particular range for the kurtosis value? any applicable range?

---

I mean Mao refinement in relion

---

map refinement*

---

Sorry, my question got lost up there - If you are using half-map-based local sharpening, should you input the ‘original’ half-maps (i.e. unsharpened and unfiltered) or filtered half-maps?

---

yes, the refinement is different than sharpening in relion

---

Can you give us a quick overview of which map should be used for further steps? Should it always be an optimally-sharpened map or should be using an unsharpened map for some steps/processes?

---

Is the b_iso on an absolute scale or just relative to the input map (which may have been previously sharpened or blurred)?

---

maps post-refinement usually need to be post-processed (aka b-factor sharpened)

---

I am sorry I could not understand whether the post processing in Relion and auto sharpening is the same thing?

---

thanks

---

is any particular range for the kurtosis value? any applicable range?

---

Should I use original half-maps to run ResovleCryoEM function first, then autoshapening; or autosharpening two half-maps individually first, then ResolveCryoEM?

---

Wouldn’t it be sharpening twice then if we perform postprocessing as well as autosharpening?

---

How does the local sharpening in phenix auto sharpen differ from other similar programs like LocalDeblur?

---

thnaks

---

New project/Set up tutorial data/Select a dataset Choose: human apo-ferritin (Density modification at 1.8 A) Hit OK once or twice to set up tutorial On left panel of Phenix GUI hit "last modified" once or twice to show the tutorial with a check mark next to it.

---

LocalDeblur is a similar procedure - the goal is to improve the map detail. The mathematics are different.

---

Which Map sharpening B factor (Å2) should we report in the publication from region or from phenix?

---

i'm sorry, but why is the density modification needed after the autosharpening?

---

It is probably good to report the original B-factor coming from the reconstruction process, and also the B-factor coming from any map improvement process (if this is reported).

---

I have an unrelated question, is the rmsd in Coot the same as the sigma level?

---

Tom’s trick for the scrolling was setting the value to 5 sigmas (rmsd option) and then use a step of 0.3 I think? (Not quite sure I caught that) this is useful I always struggle with this :’)

---

Why the density map does not cover fully the model?

---

Coot has a circular cut-off

---

The RMSD is what people often call the sigma. The pedantic among us don’t like “sigma” because it implies it’s an error level, whereas it isn’t.

---

“I have an unrelated question, is the rmsd in Coot the same as the sigma level?” — I think so

---

Run density modification ======================== Phenix GUI: Select Cryo-EM/ResolveCryoEM ... Non-default parameters: Title: apoferritin tutorial 1.8/2.2 A Half map file 1: emd_20026_half_map_1_box.ccp4 Half map file 2: emd_20026_half_map_2_box.ccp4 Resolution: 2.2 Resolution for density modification: 1.8 Seq file: seq.dat Box before analysis: False Number of processors: 4 Density modify unsharpened maps: True Final scale with FSCref: False Resolution bins: 20 Run/Run now

---

CTRL-G *in coot to Go to ... then CHAINRESID (i.e. A81)

---

You can change the radius of this circular cutoff that Dorothee mentioned

---

You can also adjust the size of that map “sphere” to whatever you feel confortable with while working on your density @Borja

---

No resolve CryoEM option in Phenix 1.17?

---

You need Phenix 1.18, or nightly build

---

Oh.

---

Can I uninstall 1.17 and reinstall 1.18 in map? Will it affect the phenix database?

---

How do you determine the resolution for DM to be 1.8?

---

When you manually re-box a map, is it easy / automatic to keep it overlaid on the original map (eg. So that a model fits into both original and re-boxed?) - I realise this is a little bit of a tangent, feel free not to address it now. Thanks :)

---

You don’t need to uninstall 1.17, you can also install 1.18 and use that. It won’t affect the database.

---

I am sorry 1.18 version of Phenix I meant.

---

This is the resolution of the reconstruction given by your software of choice before (relion etc, the resolution determined via the FSC or so)

---

@PrinceTiwari, You can do that or install 1.18.2 as well. Each version of Phenix is independent and will not mess up your project database.

---

Do we have to choose box before analysis option, if we hadn’t boxed the map like you did priorly to these maps?

---

We have tried to set things up so that boxing a map will retain the location correctly - this is a challenge in cryo-EM we discovered.

---

Yes - doing it with relion_image_handler is a pain

---

what does density modification mean? how are you changing the map and why?

---

Hi Tom, if you have a model, should you still add a sequence file? Or is the sequence extracted from the PDB? Thanks.

---

Are there tricks to formatting the seq file for complexes that are mostly RNA?

---

@Ilona: density modification uses some prior assumptions such as flat solvent regions (like in crystallography)

---

What would you recommend for a huge map for an icosahedral virus when you only need to build the asymmetric unit

---

“if you have a model, should you still add a sequence file? Or is the sequence extracted from the PDB?” — The model may not be atom-complete, while sequence is supposed to be the whole thing.

---

So does this mean that if we’re trying to use both combine_focussed_maps and density modification, we should be combining first with half-maps and then running density modification?

---

How do you choose a minimal solvent content?

---

Coming back to the huge virus map case - would you use the “box before analysis” option?

---

Try matthews for example, @Shi. Its available in ccp4

---

@lasond: It is always a good idea to choose the box before analysis option.

---

box before analysis option would minimize the size of the volume being analyzed, but would still be looking at the whole virus.

---

For combine_focused_maps I think you should density modify the separate maps and then combine them

---

If the original halfmaps are filtered, would that then negatively affect denmod?

---

Do you have an example of density distribution histograms of a typical protein? some publication/structure I can refer to?

---

@AnnaLoveland We recognize standard sequence formats like FASTA. I briefly checked the code and we will make a guess on the type of chain based on the sequence. So if you only have single letters matching nucleotides, we will most likely assume that the chain contains nucleotides.

---

and how can one obtain these density distribution histograms?

---

CryoEM density modification paper: https://www.nature.com/articles/s41592-020-0914-9

---

Thank you!

---

what about density maps of different conformations but with different density each?

---

Regarding density modification and solvent. If the interest is in seeing the solvent, even if inside the protein (not the solvent around the protein), it is no good to do density modification, or should it be fine?

---

Solvent refers to the flat (i.e disordered bulk solvent). The modification shouldn’t make the ordered solvent worse (actually should be better).

---

For density maps of different conformations do you mean half maps with different confirmations in the maps? The density modification should be done on half maps from the same reconstruction set, so I’d expect them to look fairly similar.

---

I probably missed it, but how did you decide on 1.8A for the density modification resolution?

---

Cut out map around supplied model ================================= Select Cryo-EM/Map Box Title: Box density-modified map around supplied model Map file: ResolveCryoEM_1/denmod_map.ccp4 Model file: apoferritin_chainA_rsr.pdb Output file name prefix: denmod_map_box Mask atoms: True All parameters... Search parameters/"symmetry"/OK/Ignore symmetry conflicts=True/OK Run

---

If you have a membrane protein + a detergent belt, would you also supply a model for map_box?

---

What are possible reasons for density modification not improving resolution in a noisy map?

---

@Injae: Map_box generally uses a model. Did you mean density modification?

---

Can iterative model building and density modification of original maps lead to further improvements?

---

@Dorothee I did mean map_box as it’s said to ‘cut out part of a map surrounding a model’, but Tom has clarified that now. Thanks :)

---

What happens to regions in which atoms are wrongly assigned? Does it affect the density modification outcome in that region?

---

is there a possibility to input histograms for lipids/detergents?

---

This is great, However, it is based on the assumption that we have already had a map. What if we are dealing with a unknown structure and if homologous model is quite different than the model we are dealing with. Would the method still work or do we need other procedures?

---

would you deposit a density modified map?

---

Sorry ˆˆ

---

My question got lost in-between: Regarding density modification and solvent. If the interest is in seeing the solvent, even if inside the protein (not the solvent around the protein), it is no good to do density modification, or should it be fine?

---

My personal suggestion is that apart from whatever improved maps (by sharpening, density modification, etc), also the original map only raw reconstruction. This is helpful for reproducibility and for us developers :)

---

@Andre: do you want to see ordered solvent (water molecules) or flat solvent?

---

Perfect! Thank you Tom

---

Sorry... I think I type it wrong. My question is: it is based on the assumption that we have already had a "model". What if we are dealing with a unknown structure and if homologous model is quite different than the model we are dealing with. Would the method still work or do we need other procedures?

---

@Dorothee Yes, that was the case of my question. Thank you =)

---

@Joseph, the density modification doesn’t rely on a protein model - just the original half maps and reconstruction for your protein.

---

@Paul, thank you!

---

Make copy of seq.dat with just one chain ======================================== At the bottom of the Phenix GUI select the magnifying glass Browse to open seq.dat in TextEdit or other editor Make a copy Remove all but the first sequence Save as seq_unique.dat

---

Run model-building on one monomer boxed from density modified map ================================================================= Select Cryo-EM/MapToModel Title: Quick build using boxed density modified map Map file: MapBox_2/denmod_map_box.ccp4 High-resolution: 1.8 Sequence file: seq_unique.dat Thoroughness: quick Number of processors: 4 Asymmetric map: True Run/Run now Open in Coot at end

---

If I have DNA duplex in my complex, should I provide DNA sequences in seq_unique file at this step?

---

If using the extra thorough option, at the end you get the best of all attempts or some kind of average between them?

---

Tom what is the lowest resolution that model building is generally successful in Phenix?...Or in other words ,what is it the success ratio of map_to_model according to the map resolution?

---

How about we provide a starting model?

---

also, the number of processors will only affect speed or also thoroughness in a way and also final results?

---

If you have a model for most of your complex and want to build one new density, is there a way to provide the known part of the model to restrain where the new portion will be built?

---

what about 5-6A - where you can see there are helices - will it try to build the helices?

---

Because CryoEM maps tend to have variable density, what is the best method to constrain the build to a best region?

---

where would you draw the line between this method of de-novo building and a different method, such as integrative modeling or fitting based on other data?

---

How precise of the sequence that we will need to supply?

---

Great, Many thanks!

---

I’m guessing there is some exploration of the contour to use for the model building? (Related to Dewight question)

---

what if there is no density for the loop?

---

@Ilona: Then the procedure most likely leaves a gap for that loop. Autobuild does typically not yield a 100% complete model.

---

That was the message printed in the log saying forward and reverse? (The pulchra part?)

---

sorry, I may have missed it - how is the helices directionality determined?

---

both directions are built, and the best scoring ones are kept if I got it right

---

You saw in his log there was written things like Forward: 80, Reverse:16 for example

---

thanks!

---

Does the program use any sequence-based secondary structure prediction?

---

Not currently, but we are interested in making use of this.

---

If I am using a 4.5A map and I can see the side chain density at some core regions. Can I use a map to model for the best resolution regions only?

---

@Prince, it might be best to try to build for the whole map to see what is fit - even in the less good regions there might be some mainchain fit which will help with your manual model building.

---

is there a limit to the size of the molecule it can build like that?

---

and if not size then if I have some symmetry

---

@ilona you mean the chain or the whole thing? I guess is rather related to the resolution not the size

---

Or more than resolution , ‘’map interpretability’’

---

No limit, other than map size/memory

---

ok right

---

if there are several options that get a comparable score, does it give you both?

---

But if you have symmetry then you would be using your ‘’asymetric unit’’ right? So that should involve less memory?

---

Thanks. If the map is a complex of more than one proteins then how to provide the amino acid sequence for the same?

---

@Claudia: yes, use asymmetric unit then use “apply NCS operators"

---

the fact that some regions have been built differently, or inconclusive, in various attempts (for example in the thorough option), would that help to identify for the users which regions are less ‘’reliable’’

---

?

---

Is there a way to setup the parameters for a memory intensive job on a desktop computer using the GUI then save for command line use on a server?

---

I guess you can also use the model to go back to relion and run another round of classification and refinement?

---

with the .eff file right?

---

Correct

---

Great - very helpful thanks!

---

you can put a mask on where your model is and do a focused classification

---

Remove worst-fitting residues with fit-loops ============================================ Select cry-EM/ Fit Loops Title: Replace residues 83-88 PDB file : MapToModel_3/map_to_model.pdb Map file: MapBox_2/denmod_map_box.ccp4 Sequence file: seq_unique.dat Starting residue: 83 Ending residue: 88 High-resolution limit: 1.8 Replace residues: True Run

---

If an initial model is provided which contains errors, will the program try to detect / fix these or will they be carried through to the resulting model?

---

(Sorry, for map to model, not for fix loops)

---

Thanks

---

Thanks Tom!

---

Thanks Tom.

---

Thank you!

---

Yeah. Thanks for everything!

---

It is there

---

Thanks Tom, great talk

---

At the bottom of the participants

---

Raise hand is next to yes/no

---

The first icon is ‘raise hand’.

---

Can you briefly explain the best strategy to do DM for icosahedral viral structures? Also, do we box the half-maps or the combined map in DM?

---

Thanks Tom

---

This has been very helpful for a person with almost no experience solving structures. Thank you very much!

---

You can upload supplementary maps when depositing, no need for limiting.

---

Quick question: is there a tutorial within phenix that goes through the whole process? from checking your data to getting a final model. I see many small tutorials but somehow... disconnected

---

Are we going to going to receive the recorded video today itself?

---

https://www.phenix-online.org/documentation/

---

Flow chart

---

thanks

---

I’ve seen Rosetta and Erraser tools in for modeling/refining crystallographic data in Phenix. Are these possible to use for Cryo-EM projects? Would they be used after map to model?

---

Thank you!

---

Thank you so much - this has been super helpful! :-)

---

Thanks a lot!

---

Thanks Paul and Tom

---

Thanks very much!

---

Great!!! Thanks everyone

---

Thanks you very much

---

Thank you very much