This tutorial illustrates the use of SHELXL in macromolecular refinement describing in detail the steps to follow from an initial set of coordinates to a final anisotropic model. As it is not possible to find an example covering all scenarios, some notes on aspects of wide interest such as non-crystallographic symmetry, restraints for non-standard groups and twinning are briefly illustrated in other structures. The tutorial is not intended to replace the manual and detailed explanation of the commands is avoided, it is rather a guide on how to proceed in a straightforward case, and some reference to frequent situations.

The first example pn1a has been kindly provided by Jennifer Martin (The 1.1 Å crystal structure of the neuronal acetylcholine receptor antagonist pni1 from connus pennaceus). The files required are pn1a.pdb, pn1a.hkl and optionally asn11_dis.ins.

The structure comprises 16 aminoacids, the C-terminus being modified to an amide group, and 11 solvent molecules. Asn11 is disordered.

Being a relatively small molecule, but with a protein-like structure (including 2 disulfide bridges) it is easy to follow all the typical steps in a protein refinement with SHELXL in a relatively short span of time, including anisotropic refinement, treatment of disorder and calculation of standard uncertainties.

This operation is done with the interface program SHELXPRO, which allows
(among other things) to perform some useful format transformation. Type:

> shelxpro pn1a

A menu with 26 possible options appears. Option I allows you to generate a file called p1.ins -containing instructions, restraints, atomic fractional coordinates and thermal displacement paremeters- from the pn1a.pdb file. The program will prompt you for missing information it might need, besides giving a short explanation of the operation to perform.

See page 11.1 in the manual for a commented example of a .ins file

The data file pn1a.hkl is already provided in the right format, with rfree flags set. Otherwise, options D, H (or Y) and V could be used. Exit shelxpro.

Copy pn1a.hkl to p1.hkl.

Edit file p1.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 1: rigid body refinement.

In the card TITL, it is convenient to write something informative to the purpose of the job and useful for book-keeping.

-Change CGLS 10 to L.S. 10 -1

The former would carry out 10 cycles of conjugate-gradient least-squares refinement using all reflections. The latter will perform 10 cycles of full-matrix least-squares refinement leaving out reflections flagged in the .hkl file (as -1) for the Rfree set. Full matrix refinement shows a better convergence, lower correlation, and is much slower but can be afforded at this stage given the reduced number of parameters required for rigid body refinement.

-Change SHEL 10 0.1 to SHEL 10 2.5

This sets the resolution limits being used in this job.

-Change PLAN 200 2.3 to PLAN 20 2.3

The first number defining the amount of peaks to be output from the difference electron density map. 20 is enough for such a small structure. The second figure sets a minimum distance criterium for the peaks being output to existing atoms or other peaks.

-Comment out (write REM before the lines begining with) BUMP and SWAT.

This switches off antibumping restraints (you don´t want some overlap between side chains to push the whole rigid body away when you start with a molecular replacement solution...) and bulk solvent correction (which is better to switch on latter).

-Write a line BLOC 1 N_1 > LAST somewhere before the list of atoms.

It has the effect of refining only coordinates and not displacement parameters for all atoms in the list comprising N in residue 1 (which happens to be the first atom) to whichever the last atom is.

-Write a line AFIX 6 before the list of atoms.

This signals the beginning of a rigid group. If you would want to refine several chains or domains as rigid bodies, you would have to write AFIX 0 after the last atom of a group and AFIX 6 after the first atom of the next one, and so on.

-Change the restraints generated by shelxpro for the C-terminus residue 16 to appropriate ones for an amide group:

Now you can run the first refinement job:

> shelxl p1 >p1.log&

and have a look at the files p1.log, p1.lst generated by the job.

Rigid body refinement is particularly useful when starting with a molecular replacement model. For the typical case of switching from another program to SHELXL for atomic resolution refinement, it is better to start at the next stage, namely refining all parameters from the previous model, but gradually extending the resolution in small steps until the data limit is reached. Do not forget to have the same reflections from the previous refinement in the Rfree set, if possible.

The complaint "no match for 2 atoms in DFIX" can be ignored. As the probram expects to link every aminoacid to a subsequent one, there will be 2 atoms it misses for every chain (or chain portion).

-rename file p1.hkl to p2.hkl

Edit file p2.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 2: x, y, z, u -refinement.

-Change L.S. 10 -1 to CGLS 80 -1

-Change SHEL 10 2.5 to SHEL 100 0.1

-Write a line STIR 2.5 0.02

The high resolution limit will be 2.5 to start with, decreasing in 0.02 Å steps in every refinement cycle.

-Delete REM in front of BUMP and SWAT

activates antibumping restraints and bulk solvent correction.

-Delete BLOC 1 N_1 > LAST

-Delete AFIX 6 before the list of atoms

-Move the line HKLF 4 after the comments and just on top of END.

In this way, the job titles and Rvalues will be saved throughout refinement, allowing you to keep track.

> shelxl p2 >p2.log&

have a look at the files p2.log, p2.lst generated by the job.

-copy file p2.res to p3.ins

-rename file p2.hkl to p3.hkl

Edit file p3.ins and do the following changes:

-Change TITL to TITL alpha conotoxin 3: shelxwat -n10 -s4 -m5 -u0.3 -r0.5 -w4.

SHELXWAT will call SHELXL iteratively 10 times, screen the list of residual electron density peaks from the difference map after each job, select up to 5 peaks as water molecules as long as their height is above 4 sigma, include them in the model with a starting u value of 0.3 and reject water molecules if their u value refines to 0.5 or higher.

-Change CGLS 80 -1 to CGLS 5 -1

-Write a line ISOR 0.1 O_20 > LAST

This restraint will only be used in case of anisotropic refinement of waters, later on.

-Write a line CONN 0 O_20 > LAST

To avoid including waters in the connectivity list. This helps to ensure they keep a reasonable distance to other atoms.

-At the end of the atom list, write a new residue with the coordinates of Q1, the first electron density peak

-Move the line HKLF 4 after the comments and just on top of END.

> shelxwat -n10 -s4 -m5 -u0.3 -r 0.5 -w4 p3

The file p3.ins is automatically saved to p3.bak, as it gets overwriten by p3.res at the end of each SHELXL iteration. The file p3.res contains a record of waters added per cycle. In this case, after two iterations all waters in the model had already been located.

-rename file p3.hkl to p4.hkl

Edit file p4.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 4: anisotropic refinement.

-Change CGLS 5 -1 to CGLS 20 -1

-Write a line ANIS

> shelxl p4 >p4.log&

In the file p4.lst, under the "residue reliability criteria" tables, you can see that the side-chain parameters of Asn11 fall out of the norm: a much higher residual electron density peak and hole are located close to it, and both the average and maximum displacement parameters are higher in this side chain than in any other one. Consistently, there is also an indication under the table "Principal mean square atomic displacements U". Nevertheless, the "may be split" indication can often be safely ignored, as modelling too much disordered components greatly increases the number of parameters, while not really improving the model if the disorder is not well defined. We could model disorder at this stage, but will do it at a later one, if time permits.

-rename file p4.hkl to p5.hkl

Edit file p5.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 5: H-atoms.

-Change all REM HFIX to HFIX

activates generation of geometrically placed hydrogen atoms. Hydrogen atoms on disordered residues require some manual intervention. Very high resolution refinement, when H atoms can be seen in the difference density maps uses special HFIX options. They are described in detail in the manual, sections 4-5 to 4-8.

-Move the line HKLF 4 after the comments and just on top of END.

> shelxl p5 >p5.log&

Hydrogen atoms are by default set at geometrically calculated positions respectively to the atoms they are linked to. This riding model implies that no parameters are used to refine them, so that even at medium resolution, including them in the model can be seen as a way to better account for the electron density.

-rename file p5.hkl to p6.hkl

Edit file p6.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 6: disorder.

-Write 0.6 as second parameter in the FVAR card. This gives free variable 2 a start value

-Move the line HKLF 4 after the comments and just on top of END.

> shelxl p6 >p6.log&

Disorder can be limited to part of the side-chain as well. In that case, just write the alternative positions for disordered atoms under part 1 and part 2 respectively.

-rename file p6.hkl to p7.hkl

Edit file p7.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 7: Final refinement against all data.

-Change CGLS 20 -1 to CGLS 20

-Move the line HKLF 4 after the comments and just on top of END.

> shelxl p7 >p7.log&

You can use shelxpro to generate some plots useful to check your structure. Try options P, T and R:

> shelxpro p7

-copy file p7.res to p8.ins

-rename file p7.hkl to p8.hkl

Edit file p8.ins and do the following changes:

-Change card TITL to TITL alpha conotoxin 8: standard uncertainties.

-Change CGLS 20 to L.S. 1

-Write a line DAMP 0 0

This will prevent parameter shift despite removing all restraints.

-Delete all restraints: lines begining with SIMU, DELU, ISOR, BUMP, DFIX, DANG, CHIV, FLAT

Restraints lead to an underestimation of s.u.

-Move the line HKLF 4 after the comments and just on top of END.

> shelxl p8 >p8.log&

Notice there is still one restraint generated by the program to fix y, P21 being a polar spacegroup.

You can use shelxpro to plot esd´s against B values. Try option E:

> shelxpro p8