Re: [ccp4bb] How to convince oneself (and others) of ligand presence
Dear all First of all, thanks for the replies here and off-list. I reprocessed the data to 2 A, and modeled two copies of the ligand in each of the two protein NCS molecules, without using NCS restraints for refinement. I also calculated an SA omit map and a polder map. In the end, the ligand molecules have a relatively low B factor and CC of 0.8 and 0.9. The fact that these two ligands are in the same place (active site) in both NCS copies suggests they are not some random noise. However, I also could model two more of the ligand outside of the active site, also with low B factor and CC > 0.8. Can I call these 'adventitious' molecules? Thanks. Mohamed On Thu, 23 Feb 2017 13:56:05 +, Mohamed Noor wrote: >Dear all > >I think I have a ligand (substrate) placed in the active site of my model >correctly. The CC is 0.785, B factor of 60 A^2 which is roughly similar to the >neighboring residues and there was no ligand in my search model*. The data >extends to 2.2 A, although I think I can get something higher with manual >processing**. How can I be really sure that it is the ligand that I think it >is? And assuming you are a reviewer, what would you expect to see in a >manuscript? > >* The unliganded model was obtained by molecular replacement using a search >model that did not have the ligand. I then used this model to simply do one >cycle of rigid-body refinement and another one cycle of coordinate/ADP >refinement. > >** The data was auto-processed during data collection using xia2/XDS. I have a >dozen of datasets so I am just taking the autoprocessed ones and see if my >ligand is there. > >Thanks. >Mohamed
Re: [ccp4bb] How to convince oneself (and others) of ligand presence
PanDDA tries to arrive at an objective score metric for presence: http://biorxiv.org/content/early/2016/09/05/073411 But in the end, you still have to apply all the scientific rigour that Bernard mentions. On 23/02/2017 13:56, Mohamed Noor wrote: Dear all I think I have a ligand (substrate) placed in the active site of my model correctly. The CC is 0.785, B factor of 60 A^2 which is roughly similar to the neighboring residues and there was no ligand in my search model*. The data extends to 2.2 A, although I think I can get something higher with manual processing**. How can I be really sure that it is the ligand that I think it is? And assuming you are a reviewer, what would you expect to see in a manuscript? * The unliganded model was obtained by molecular replacement using a search model that did not have the ligand. I then used this model to simply do one cycle of rigid-body refinement and another one cycle of coordinate/ADP refinement. ** The data was auto-processed during data collection using xia2/XDS. I have a dozen of datasets so I am just taking the autoprocessed ones and see if my ligand is there. Thanks. Mohamed
Re: [ccp4bb] How to convince oneself (and others) of ligand presence
At the risk of being redundant, I can only restate what I and many others have posted before: No simple answer or single universal statistic for your justified question or concern exists. The crucial question to ask first is what hypothesis or claims is your (ligand) model at that specific location supposed to support? A specific strong claim or hypothesis demands the corresponding degree of solid evidence and high plausibility of the model. If the pose of your ligand is a crucial part of your claim or hypothesis (and a small molecule at a specific location in a specific conformation is generally a pretty strong claim), both the electron density and environment, as well as the plausibility given prior knowledge need to be convincing. Also recall that the RSCC is a crude measure generally applied to the entirety of your ligand. So you may have a somewhat lower RSCC in certain situations like a piece of a ligand being floppy, or, with very weak density, suffer from solvent intrusion in omit maps (although in this context I caution against a variety of tricks and methods if you you do not exactly know what you are doing). Even in such cases, the rest needs to make sense. Really bad ligands that score high in the 'bad' lists generally exhibit multiple features like poor density fit, no convincing binding site contacts, collisions, and poor stereochemistry. Your 0.8 RSCC is not super good, but not bad either. You need to make the case to the reviewers based on your specific claim/case. It is important not mistake this disclaimer for a postmodern argument justifying ‘Anything goes’. Again, if one’s claim is to support a certain mechanistic detail or ligand pose, one better cough up some convincing, properly generated and adequately contoured (difference omit) electron density and a model that is not delusional in view of basic stereochemistry. A problem I see is as a consequence of the 'positive results only' philosophy, in order to make the high impact list, it is almost imperative to make powerful statements and associated exaggerated claims. The moment you voice the slightest doubt (i.e. meaning you are in fact honest about what you can say with a given certainty and what is doubtful), some reviewer (usually the 3rd one) will make a clueless comment about R-merge and it is Acta F. :-) Best, BR - Bernhard Rupp b...@ruppweb.org hofkristall...@gmail.com http://www.ruppweb.org/ - The hard part about playing chicken is to know when to flinch - On Thu, Feb 23, 2017 at 5:56 AM, Mohamed Noor wrote: > Dear all > > I think I have a ligand (substrate) placed in the active site of my model > correctly. The CC is 0.785, B factor of 60 A^2 which is roughly similar to > the neighboring residues and there was no ligand in my search model*. The > data extends to 2.2 A, although I think I can get something higher with > manual processing**. How can I be really sure that it is the ligand that I > think it is? And assuming you are a reviewer, what would you expect to see > in a manuscript? > > * The unliganded model was obtained by molecular replacement using a > search model that did not have the ligand. I then used this model to simply > do one cycle of rigid-body refinement and another one cycle of > coordinate/ADP refinement. > > ** The data was auto-processed during data collection using xia2/XDS. I > have a dozen of datasets so I am just taking the autoprocessed ones and see > if my ligand is there. > > Thanks. > Mohamed >
Re: [ccp4bb] How to convince oneself (and others) of ligand presence
Well - I certainly would do more refinement - if something is bound there will be other changes in solvent etc. then search the difference map for peaks/ find ligand etc, and see what it shows.. If your original model and this one are isomorphous you can also check the Fobs-lig -Fobs-nolig Phi-nolig but that may not be necessary eleanor On 23 February 2017 at 13:56, Mohamed Noor wrote: > Dear all > > I think I have a ligand (substrate) placed in the active site of my model > correctly. The CC is 0.785, B factor of 60 A^2 which is roughly similar to > the neighboring residues and there was no ligand in my search model*. The > data extends to 2.2 A, although I think I can get something higher with > manual processing**. How can I be really sure that it is the ligand that I > think it is? And assuming you are a reviewer, what would you expect to see > in a manuscript? > > * The unliganded model was obtained by molecular replacement using a > search model that did not have the ligand. I then used this model to simply > do one cycle of rigid-body refinement and another one cycle of > coordinate/ADP refinement. > > ** The data was auto-processed during data collection using xia2/XDS. I > have a dozen of datasets so I am just taking the autoprocessed ones and see > if my ligand is there. > > Thanks. > Mohamed >
[ccp4bb] How to convince oneself (and others) of ligand presence
Dear all I think I have a ligand (substrate) placed in the active site of my model correctly. The CC is 0.785, B factor of 60 A^2 which is roughly similar to the neighboring residues and there was no ligand in my search model*. The data extends to 2.2 A, although I think I can get something higher with manual processing**. How can I be really sure that it is the ligand that I think it is? And assuming you are a reviewer, what would you expect to see in a manuscript? * The unliganded model was obtained by molecular replacement using a search model that did not have the ligand. I then used this model to simply do one cycle of rigid-body refinement and another one cycle of coordinate/ADP refinement. ** The data was auto-processed during data collection using xia2/XDS. I have a dozen of datasets so I am just taking the autoprocessed ones and see if my ligand is there. Thanks. Mohamed