It's a bit tricky refining occupancies at that resolution, to be able to do it properly you would need better resolution, as the occupancy is otherwise heavily coupled to the B-factor.. However, if you do believe that you do have a lower occupancy of your ligands, you could try to let Phenix automatically refine the occupancy of the ligands for you. To do so, set the occupancy to lower than 1 for the ligands, and click the option for occupancy refinement. I would then be careful when inspecting the outcome.

Thanks for the prompt reply. There is no specific reason for me to believe that there is less occupancy for the ligand or for any conformation of the residues in the protein. My B-factor is unusually high and i was wondering if i could reduce it by lowering the occupancy. If resolution is an issue i would rather leave it (B-factor) at that.

I agree with Ronnie re "a bit tricky refining occupancies at this resolution". You truly need more data to accurately model the thermal parameters. If you look through your coordinate file do you notice a region of atoms that have elevated B factors as compared to the mean? In the past, I have noticed that some solvent accessible residue side chains have little to no density and therefore have elevated B-factors (>100 A^2/(8*pi^2)). Removing the carboxylates of some offending Asps and Glus was required for this particular example, which decreased the overall B-factor without messing with the occupancy. Do you have a low Rmerge? Mosaicity? How is your completeness? Have you modeled the solvent yet? Did you fully model your structure prior to placement of the ligands? Was TLS applied?

If you generate a Composite Omit Map you may be able to pin-point regions of your model that are ill-fitted to the density. This would be a good starting point to make your model better and assess what needs to be happen to complete the refinement. And, as Ronnie mentioned above, be careful when analyzing the output. You can very easily bias your model here.

Well, the answer in most of the cases is "you cannot refine atomic displacement parameters and occupancy". Both parameters are tightly correlated at low to mid 2-Theta angels. One can try to disentangle them at higher 2-Theta angles but you need really an exceptional quality of diffraction data. Something you apparently do not have. In order to be absolutely sure that you do not produce rubbish results you would have to measure your sample at very low T ( a few K would do) to reduce the dynamic disorder in your crystal.  Neutron diffraction would be of course the best option.

I would never reduce the occupancy of protein atoms below 100% unless that specific protein chain is not participating in any crystal contacts. In almost every case you know those protein atoms have to be (somewhere) there because the molecule generates that crystal itself.

An average B-factor of 56 is not extremely high at 2.3A imho. Maybe your refinement wasn't restrained enough and that pushed your B-factors a little higher than you expected? How are the average B-factors of the backbone up to the c-beta or c-gamma atoms?

Wilson B and average B are two totally different things. Have you tried squaring up your model structure factors and calculate a Wilson B from those? That Wilson B has quite a few raw assumptions in it and is pretty sensitive to data quality/processing (overloads, beam stop shadow, ice rings, etc.). Do you have many missing atoms in the structure (disordered loops) ? If you suspect an issue here with the Wilson B the first thing to do is go back to the frames. Don't be  afraid to spend a day or a week over your data if it's troubling you. Data not interpreted/processed right can cause month worth of trouble later.

For the ligands, I wouldn't worry too much unless their B-factor is drastically higher than the closest protein atoms and/or the maps show signs of model errors. If that's the case, I'd recommend to set the ligand occupancy to a realistic value (e.g. 85%?) and leave it fixed during refinement and monitor the B-factors of the ligand and it's environment. A ligand fully enclosed in a binding pocket should have B factors more similar to the surrounding protein atoms than a ligand lying flat on a surface, obviously.  How do other non-water solvent atoms behave in the refinement? (if you have any).

Congrats!! You solved and (almost!!) fully refined a protein structure! It is a very good habit to look into your work and to question your model and to ask for help when unsure. I am certain you are just a mosquito sneeze away from your final model. B-factors can be mean but I honestly believe you are on the good side of trouble, according to your description.

Hi Vijay,

you already received good answers, so shortly to sum up:

- As you already know from the comments above "you cannot refine atomic displacement parameters and occupancy AT THE SAME TIME", so you should "determine" one parameter and leave it fixed. 

- Your B-factor is not suspiciously high at this resolution (Don´t worry too much about high B-factor values)

- Check the B-factor values for residues arround your ligand

However, I wanted to add something:

First of all you have to convince yourself that the ligand is present (is there density?, does the ligand fit into this density?, does it make sense according to ligand coordination, chemistry and the ligand enviroment?, etc ....) and if you are sure it shouldn´t be a problem to convince a reviewer, because most of the reviewer doesn´t care if the ligand is present with an occupancy of 1.00 or 0.80 (or soemthing), so you shouldn´t spend too much time about this!

However, I experienced a situation, where I was forced to "give an exact occupancy value" for all atoms of my ligands by a reviewer. My resolution was 1.8 A and I started a marathon of refinements (which was suggested by crystallography experts) in order to determine "exact values" for both occupancies and B-factors SEPARATELY .....  Briefly, I started X refinements with different values for the B-factor (e.g. my 10, 20, 30, 40 and so on) until I reached convergence. I selected refined structures with similar (low) R-values and I checked the B-factors of these structures, which clustered arround a value (e.g. 10 structures with B-factors of 30-40) , so I took the average value of this "clustered values" (e.g. 35) and left the B-factor fixed. Afterwards the same procedure was done with the occupancy. I don´t know how much sense it makes to do this in your case, but I just wanted to tell what I have done when I was asked about the occupancy of my ligands. Maybe it is useful.

Cheers,

Aleks

Thanks Stefan for the encouraging comments.Yes, i do have missing residues in my protein. Around 15 residues in the c-terminus are missing and i am not able to find any electron density and this 15 residues are not there in my MR model as well. Are there ways to find electron density? Can i use alternative MR tools to check this? There is enough space in that region to model 15 residues.

As Alexander commented the B factor for the residues surrounding the ligands were low. Based on this i have removed some ligands and did few more rounds of manual refinement. Now the average b-factor is 48. I think this is an accepted statistic. Thanks for all your suggestions.

Apart from wwpdb validation tool what are the other validation tools that i need to use to ensure that if any gross errors or outliers are there in the structure model?