Hi Jianghai Zhu,

thanks for your questions!

<>I have a structure refined to 2.5 A. After simulated annealing (start temp = 5000, cool rate = 25, final temp = 100, include TLS refinement and NCS restraint), my R and Rfree are 24.6% and 28.7% respectively. I then turned off the SA in the generated def file and carried on the combined refinement. My R and Rfree went up to 25% and 29.2% respectively. Why did the Rfree go up after refinement?

The difference between 28.7% and 29.2% is not large (=0.5%). If you run a set of identical refinement jobs with different random seed you will get the distribution of final R-factors ranging within 0.5%. So I would not worry too much about this.
However, I would try different refinement strategy, for example:
1) you do SA refinement first;
2) then you do regular refinement + TLS.

The RMSD of bonds and angles after either SA or combined refinement are about 0.006 and 0.851 respectively. Don't you guys think the geometry restraints are too tight?

At 2.5A resolution I wouldn't worry about these values. The resolution is low enough to keep the geometry tight.

How do I change the weight in PHENIX?

However, if you want you can play with the target weights. The total target function in phenix.refine for the coordinates refinement is :
T = wxc_scale * wxc * Exray + wc * Egeom, where wxc is determined automatically and wxc_scale is a parameter that the user can modify. So, if you want to make the X-ray term stronger, then obviously you need to increase the wxc_scale value. By default it is 0.5 (see parameters file under the scope refinement.target_weights). Try several wxc_scale and pick the refinement outcome you like best.

Let us know if you have more problems / questions.

Cheers,
Pavel.