Previous: What does each calibration file do?
Next: How to convert KappaCCD images for use with APEX2 and SAINT

Using EVALCCD for data reduction

Introduction

Most current KappaCCD users are used to the HKL software for the integration of the data. Now EvalCCD is presented as an alternative.

The philosophies behind the HKL software is quite different from the principles of EvalCCD. To make optimal use of EvalCCD, even the cell determination should be adapted.

This document will introduce the EvalCCD users to some of the philosophies. We will make a quick walk through a data collection and integration here, making reference to manual pages of the individual programs for details.

Existing KappaCCD users should note that they should use an "easy" crystal for their first excercises with EvalCCD, and not a crystal that they have not been able to treat before.

Background of EvalCCD

Traditionally, integration programs for X-ray area detectors have come from the protein crystallography world. This means that these programs have many tools to deal with the difficulties of protein data integration and synchrotron sites: nonlinear decay of crystals, unstable beam intensity, imperfections in the goniostat hardware. Many parameters dealing with these problems are refined during routine data reduction. Small molecule crystals on the KappaCCD mostly do not suffer from these problems.

Small molecule crystals, however, have difficulties of their own. For example: twinning and incommensurate modulations. Data need to be determined to much higher diffraction angles. And a home source has its own difficulty too: the characteristic Mo, Ag or Cu radiation is not monochromatic, but contains two distinct lines K-alpha1 and K-alpha2 that are clearly separated at high diffraction angles.

The traditional integration software has incomplete answers to most of these small molecule crystallography problems. Sometimes they can be retrofitted with new correction procedures, but often the result is not optimal.

EvalCCD was developed with a completely different mindset: it should be able to cope with difficult small molecule crystals. And because even the smallest deviation from a predicted value (e.g. a differently shaped reflection) could be an indication of twinning or a modulation, we do not want to allow parameters to vary if we know they should be constant. Hence, we do not "derive" the shape and position of reflections from the images, but we "predict" them from goniostat and crystal constants. This prediction method was developed by Dr. A.J.M. Duisenberg of the Bijvoet Center for Biomolecular Research at Utrecht University in the Netherlands. Its algorithm, known under the name "eval14", is described in his PhD Thesis (Reflections on Area Detectors, Utrecht University, 1998), and in a more recent paper (An intensity evaluation method: EVAL-14, A.J.M. Duisenberg, L.M.J. Kroon-Batenburg and A.M.M. Schreurs, J.Appl.Cryst 36 (2003) p.220-229).

A consequence of this invariability of constants is that the constants should be accurately known before the data collection is started. To obtain accurate goniostat constants, a calibration procedure needs to be followed using the "perfect" AMBI crystal. To obtain accurate unit cell constants, a specially developed phi/chi procedure is carried out. And to obtain a macroscopic shape and orientation of the crystal a rough face-indexing procedure is done.

Armed with all this prior knowledge, the data can be integrated with great precision, even when the crystal shows small-molecule type problems.

Getting started using EvalCCD

Collect can use either the HKL software (denzo/scalepack/xdisp) for integration, or the dirax/view/EvalCCD programs from Utrecht University. The default configuration is to use the HKL programs. Before using any of the EvalCCD related programs, please make sure you use the default_eval() function at the top of your configuration file. If you do not have denzo installed at all, add the associated configuration parameter "denzo_installed=0".

To activate the EvalCCD interface in the "collect" suite, a special license to the evalccd module is required. This license can be purchased at your local Nonius representative, and can be obtained using the nlicense program.

Both with the HKL software and with the EvalCCD software, the "collect" program suite only contains an interface to make the usage of the original programs available. The original programs are distributed separately. Before you can use the EvalCCD programs, you will either need to download the evalccd package from the Nonius website, or buy the latest Nonius software CD/ROM.

Before using the EvalCCD software, you also need to make sure your detector has undergone all possible calibrations. Until at least January 2000, systems were delivered only with those calibration files required for running the HKL software and possibly some "collect" tools. Maintaining a proper system calibration is essential if one wants to get the most power out of the EvalCCD software.

From crystal to images

The supergui program can automatically guide you through the complete process of data collection and integration.

Cell determination

Since in EvalCCD, the unit cell is not refined during the integration, an accurate orientation matrix is required before the integration is commenced. For this goal the phi/chi procedure was developed (A.J.M. Duisenberg, R.W.W. Hooft, A.M.M. Schreurs and J. Kroon, Accurate cells from area-detector images, J. Appl. Cryst. (2000) 33, 893-898) Please use the phi/chi experiment to determine your unit cell whenever possible.

It is possible to use a standard set of rotation images to find a unit cell using dirax (using the command line tool rotindex), but since the tolerances needs to be relaxed for such a measurement it will be much more difficult to resolve twins and detect other indexing problems.

If you have multiple unit cells (e.g. a twin or a fragmented crystal), you will need to run the ndirax program, or in difficult cases you may need to run "dirax" program manually. You should write ".rmat" files associated with each of the relevant unit cells you can locate.

The integration procedure will use the primitive cell that is normally found by dirax. If there is a centered cell with higher symmetry, the rmatrix program should be used to create a .rmat file containing the centering transformation. If this .rmat file is used for the integration, the final data set will already be based on the centered cell.

The program comparecell is available to find relationships between two unit cells you have found. This program can also be useful to find non-trivial lattice transformations that could be the cause of twinning.

Finally, there is a program ntrans that can be used to perform some transformations on unit cells, e.g. if you know that there is a twin lattice but you are not able to locate it using dirax.

Measure crystal

The prediction of the reflection outline in Eval14 relies on a rough description of the crystal: the crystal is either described as an oriented block (which means 6 parameters: 3 sizes and 3 orientation angles) or as a sphere (only one parameter: the diameter). The procedure will automatically use a face index description (if available) to derive the best crystal description. It is therefore a good idea to use the Orientation experiment in collect to make a face index description of the crystal.

If absorption correction will not be required, a quite rough description will suffice to get an adequate description into EvalCCD.

Please note that particularly for strong absorbing crystals, the possibility to perform analytical absorption correction is often advantageous: if you have the possibility to make an accurate face-index description, please take the time to do this.

Study images

The findresolution program available under "Guess diffraction limit" does more than just showing you a table of how well your crystal diffracts. It also stores the information in the so-called "hints" file, such that the collect program will be able to use better defaults. Please run this program before commencing strategy calculations if possible.

Data collection

Just follow the data collection experiment to get images. Never underestimate the importance of data redundancy, especially when dealing with absorption or fragmented crystals. EvalCCD can handle finely-sliced data as well as widely-sliced: collect will choose whichever strategy is the most suitable for your crystal.

From images to reflections

The integration is performed in two main parts: in the first part, a "shoe box" is made for each reflection: the relevant part is cut out of a number of successive images, and stored in a "shoe box file". In the second part of the integration, the "shoe boxes" will be integrated into reflection intensities.

Make shoeboxes

This is the first major part of the integation. It is performed by the "Make shoeboxes" button, which runs the makeshoe program. All shoe box files will be created in subdirectories of the data directory.

Unless instructed otherwise, this procedure will show you every tenth image, together with the shoe box outlines. This way you can see whether the reflections are in the shoe boxes.

Tune crystal parameters

Runs the shoe box integration in "interactive" mode. The program that does this is called makextalevc. You will be able to see each reflection after it is integrated, and modify the parameters of the integration procedure to optimize the results.

Integrate shoeboxes

Runs the shoe box integration in "batch" mode. This button runs the evalall program. If you are integrating a crystal with more than one entity, you need to run evalall on each of the shoe box directories.

Post-process data

This runs the nanny program to analyse the data. This program will allow you to adapt the error model, and include and reject groups of reflections based on selection flags created by the integration.

This program can write a "shelx hklf4" file for a single entity, or a "shelx hklf5" file for a two-entity crystal.

Storing images for later perusal

Two issues:

Since the "EvalCCD" procedure relies on accurate calibration, make sure to write the contents of your calibration directory to the CD/ROM along with the data. The calibration of the machine may change before you want to reprocess the data. Also, your data may be processed on another computer that does not have your calibration data at hand.
Since all programs from the "collect" and "EvalCCD" program suites can read compressed images, it is safe to write a CD containing the compressed images. You will not be required to copy all images back to hard disk once you are ready to reprocess them.

Previous: What does each calibration file do?
Next: How to convert KappaCCD images for use with APEX2 and SAINT