1) Check for the Sigma level of the outer shell the density map, some times you need to consider the higher resolution cutoff i.e. 2.6 to 2.7 A, and model again.

2) Manually check and add the water molecules in an unmodelled density map and do refinement. (trial and error).

3)Redo the scaling and merging process to get the improved statistical .mtz file.

4) Improve the crystal quality to get high-resolution data. Check for overall completeness.

5) Adjust the energy and exposure time during the data collection.

@suparna

Thanks for the suggestions. So I have tried most of the suggested points. I have added water as well as buffer components however the rfree and rfactor difference remains the same (~ 8%).

So is the difference acceptable as I have restrained the bfactor (I am prediciting the higher difference is because of this) as I wasn't getting density for a loop (and post restraining I got the density) which is important for my structure.

As Suparna G suggested, may be you have to try them all and see which works, may be getting better diffraction data will help you in getting better resolution for loop region (if you just need to reduce the difference then you may even solve at lower resolution cut-off from 2.6 to 2.8 or even 3A). Many other information (Space group used, scaling/ merging stats, similarity between the template used for MR, crystal packing, etc.) you may provide to get the best solutions.

You might also share the question on CCP4 mailing list, where the experts will give you better solutions to your crystallography related question.

Hi Shubhashish,

Did you solve your problem?

There are several questions and suggestions on how to reduce difference between R free and R factor.

1. What was the space group of your crystal?

2. How many chains are in the model?

3. Did you use TLS groups?

Is it possible that you processed your data in higher symmetry that your crystal belongs to? I would suggest to check that carefully.

If you have several chains, you can use non-crystallographic symmetry (NCS) restrains in Refmac. Sometimes it helps.

Using TLS correction helps to reduce overall R factor, which will help to see more details in your maps.

The refinement is an iterative process. First I fix the big problems (moving main chain into the density), and I am not trying to fix all problems at once. If some places are not clear, I do my best, but I do not spend too much time. I put the edited model in new rounds of refinement.

Usually, I check all residues at least twice. My steps:

1. The model from the molecular refinement goes in Refmac. Several rounds 10 cycles each. I check if my numbers are changing. I will setup manual weight and keep rmsd for bonds around 0.006. I will set up my own geometry restrains: bonds length 1.1, angles 1.2, torsion angles 2.7 - 3.0, and B-factors 0.8.

2. When I see that refinement converged I will run ARP/wARP and generate waters. And refine the model for several rounds to make sure that it converged.

3. First check of the model in Coot. Starting with the residue 1 (or whatever is the first residue) and one by one checking the entire model (pushing the space bar in Coot). I work in stereo that makes this process much faster. I do not need to rotate the model too much. At this point I would move main chain, side chain and add new waters manually.

4. The updated model will be refined several rounds. At this point I would check geometry. Running MolProbity (server) and checking all close contacts, main chain and side chain outliers.

5. I check the R factor and R free. I use tricks like NCS restrains, generate TLS groups. And I will refine the model with the adjusted parameters. Which TLS group refinement, I reduce the weight by 50% to keep the proper geometry.

6. I check again the whole model, and I would add new waters, alternative conformations. I would check 2Fo-Fc maps at levels below 1.0 sigma. Add partial (occupancies below 1.0) waters, waters with the proper distances to protein but placed into density below 1.0 sigma level. While checking the model, I would adjust water occupancies and generate alternative waters (peanut shaped density).

7. I would place the model in next rounds of the refinement. Run MolProbity and see if the model has problems.

If after all these steps the R free is still 8% higher than R factor, I would check my data. Twinning, pseudo-symmetry and concidering to collect a new data.

Hope this helps.

Best,

George.

Pankaj Thapa Thanks for the suggestion.

George Minasov Yes I have solved the structure with a lower rfactor and rfree. However, thanks a lot for your suggestions, I would definitely keep these in mind while solving the structures further.

Hi Shubhashish,

I am glad that you solved your problem.

Sometimes we are not so lucky.

Best wishes,

George.