Dear Wout, A suggestion from the inventor of the FRC/FSC: please read the discussions on using non-linear refinement procedures to optimise linear metrics beyond their defined validity range in: "Information: to Harvest, to Have and to Hold", by Marin van Heel & Michael Schatz. (https://arxiv.org/abs/2009.03223) Groet, Marin
On Thu, Sep 10, 2020 at 4:26 PM Oosterheert, W. (Wout) <[email protected]> wrote: > Dear all, > > > > Thank you very much for the insightful comments. Intuitively I would also > say that using a postprocessed map that is optimized to look more like a > protein model, shouldn’t be used to real-space refine the model. Therefore, > I’ll keep performing my real-space refinements in the experimental maps. > That being said, I would still very much recommend DeepEMhancer for model > building and for map visualization in ChimeraX; in my experience, the maps > look clean and less anisotropic. > > > > Best, > > Wout > > > > *Wout Oosterheert; PhD Candidate; Crystal and Structural Chemistry; > Bijvoet Center for Biomolecular Research; Utrecht University; The > Netherlands* > > > > *From:* Collaborative Computational Project in Electron cryo-Microscopy < > [email protected]> *On Behalf Of *Schmid, Michael F. > *Sent:* Thursday, 10 September 2020 20:09 > *To:* [email protected] > *Subject:* Re: [ccpem] Refinements against DeepEMhancer maps > > > > Hi- > > I like to remind that what we are doing in EM is not refinement, in the > crystallographic sense, but model optimization into whatever map we are > using. > > And as for the dependence on resolution, a difference of 0.2 Å (I won’t > call it an error, since highly refined coordinates can vary this much, at a > 2 sigma, 5%, frequency) will not be detectable in the FSC at 2.5 Å, but > will start to give bad correlations at high resolution in a 1.2 Å map. The > model has to be more correct at high resolution to give a good FSC than it > has to be at low resolution, but this is obvious. > > Mike > > > > *From: *Collaborative Computational Project in Electron cryo-Microscopy < > [email protected]> on behalf of "Pintilie, Greg" < > [email protected]> > *Reply-To: *"Pintilie, Greg" <[email protected]> > *Date: *Thursday, September 10, 2020 at 10:52 AM > *To: *"[email protected]" <[email protected]> > *Subject: *Re: [ccpem] Refinements against DeepEMhancer maps > > > > > > I fully second Marta's comment; why not use the full information to build > a more accurate model. Just watch out for over-sharpening, it's typically > easy to detect visually in the form of disconnected densities; unless you > have a 1.22Å map, then disconnected densities are OK :) > > > > Whether to use in refinement is a trickier question, because you can get > stuck in local minima easier when there is more detail. It could help to > start refining in the less detailed map, and then move to the more detailed > map. Use manual adjustments in Coot/ISOLDE to get out of local minima... > > > > Lastly, if we are using a real-space score, but get lower scores when > there is more detail in real-space, then I suppose we might have some > questions about whether that's a good score to use? > > > > Kindly, > > > > Greg > > > > > ------------------------------ > > *From:* Collaborative Computational Project in Electron cryo-Microscopy < > [email protected]> on behalf of Marta Martinez <[email protected]> > *Sent:* Thursday, September 10, 2020 12:59 PM > *To:* [email protected] <[email protected]> > *Subject:* Re: [ccpem] Refinements against DeepEMhancer maps > > > > Hi all, > > I have also used DeepEMhancer, as well as other sharpening methods, > and I usually consider sharpened maps as a good help for tracing in > Coot. They contribute to solve quite a few ambiguities. Normally, I > have both the experimental map (coming from Relion in your case) as > well as the sharpened map (one or several) aligned together to > constantly assess the reliability of the improvements of tracing that > I can get with the sharpened maps, especially in controversial areas. > > Commonly, I perform the tracing in Coot based on the sharpened maps. > Regarding refinement, in automatic refinement steps, however, I only > consider the experimental map and all my validation scores are > referred to that map (higher CC is normal). Nevertheless, small > corrections of manual refinement are also performed in Coot and then > the sharpened maps are again a good help. As you might have > experienced, several rounds of automatic and manual refinement are > common to constantly improve the validation scores (referred to the > geometry of the molecule and map-model correlation). Having > experimental and sharpened maps aligned allows you to assess the > reliability of the improvements in tracing that you can get with the > sharpened map. > > I hope this help. > > Best regards: > > MARTA > > > Quoting "Oosterheert, W. (Wout)" <[email protected]>: > > > Dear all, > > > > I’ve been playing around with map postprocessing using DeepEMhancer, > > and I’m quite impressed by the results, especially for datasets that > > suffer from preferred-orientation. > > > > I recently obtained a 3.4 Å, gold-standard refined map using Relion, > > and subjected the unfiltered half-maps to postprocessing using > > DeepEMhancer. I then tried to refine my model against the > > DeepEMhancer-postprocessed map using Phenix Real-space refine. The > > model looks fine after refinement, but I noticed some strange values > > in the validation statistics: > > > > * The model vs. data CC is only 0.73, whereas the CC is 0.84 > > when I refine against the Relion-postprocessed map. > > * The map to model FSC curve tails off very slowly, FSC = 0.5 is > > at 3.2 Å, and FSC = 0.3 at 2.2 Å. Is this type of behaviour expected? > > > > In the bioRxiv manuscript > > (https://www.biorxiv.org/content/10.1101/2020.06.12.148296v3), > > manual model building using Coot is performed in the DeepEMhancer > > maps, but refinements are not mentioned. So my question is: > > can/should I run refinements against DeepEMhancer-postprocessed > > maps, or should I only use these maps to guide model building in Coot? > > > > Thanks, > > Wout > > > > Wout Oosterheert; PhD Candidate; Crystal and Structural Chemistry; > > Bijvoet Center for Biomolecular Research; Utrecht University; The > > Netherlands > > > > > > ######################################################################## > > > > To unsubscribe from the CCPEM list, click the following link: > > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1 > > > > This message was issued to members of www.jiscmail.ac.uk/CCPEM, a > > mailing list hosted by www.jiscmail.ac.uk, terms & conditions are > > available at https://www.jiscmail.ac.uk/policyandsecurity/ > > > -- > Marta Martinez Gonzalez > > Biocomputing Unit (Lab B13) > National Center for Biotechnology-CSIC > Darwin, 3. Campus de la Universidad Autonoma de Madrid > 28049 Madrid. Spain > Tel:+34 915854510 Fax:+34 913720112 > E-mail: [email protected] > > ######################################################################## > > To unsubscribe from the CCPEM list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1 > > This message was issued to members of www.jiscmail.ac.uk/CCPEM, a mailing > list hosted by www.jiscmail.ac.uk, terms & conditions are available at > https://www.jiscmail.ac.uk/policyandsecurity/ > > > ------------------------------ > > To unsubscribe from the CCPEM list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1 > > > ------------------------------ > > To unsubscribe from the CCPEM list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1 > > ------------------------------ > > To unsubscribe from the CCPEM list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1 > ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
