| DPS Home | Supported Detectors | Download DPS | Program Details | FAQ | Software history | References | Mailing list |

DPS_scale

Frozen crystals of proteins, nucleic acids or other biological macromolecules often have mosaic spreads comparable to the maximum useful oscillation angles. It is therefor necessary to use scaling methods that are independent on the exclusive use of full reflections. This can be accomplished by either summing intensities of partial reflections from consecutive frames (  method 1 ) or by correcting the intensities of partial reflections based on a model for the reflection partiality ( method 2 ).  The latter method has been proven to be more stable and versatile, allowing scaling of data sets with incompletely measured reflections and low redundancy. Data from low symmetry crystals or data collected over small rotation ranges generally has low redundancy. Method 2 is also less sensitive to dramatic changes in the true scale factor between consecutive frames, e.g. due to beam dumps and refills at a synchrotron source. 

The full scaling process involves 4 different steps. Before the scale parameters can be calculated, the reflections useful for scaling have to be determined and sorted. Any reflections listed to be rejected in an user supplied file are eliminated. Any re-indexing of frames also happens at this step. Re-indexing might be necessary to bring the frames of 2 separately indexed and integrated batches in the same orientation or to simply change axes to bring the crystal symmetry in a standard setting. Finally the reflections are sorted and written to the file SORTED_REF. 

Scaling means the least squares determination of such parameters as the linear scale factors, B-factors, mosaic spreads, cell parameters and setting angles. Constrains and restrains can be applied for these parameters. Restraining least squares parameters means introducing a parameter "restrain", such that the following function will be added to the target for the least squares minimization:

( parameter(i) - parameter(i+1) ) ** 2 / restrain**2, 

where parameter(i) and parameter(i+1) are the least squares parameters (for example, scale factors or B-factors) of the i-th and (i+1)-th frames correspondingly. By default no restraints will be applied.
Constraining least squares parameters means forcing a parameter to follow a given mathematical function, in the case of DPS scale a fourth order polynomial. This function is currently under development and should not yet be used. 

After merging the frames, DPS creates the file SORTED_REF.

Each line of this file contains

The value of flag1 is decided based upon the mosaicity and orientation matrix after integration.
flag1 Description.
0 Reflection not predicted by DPS_scale.
Reflection entering the Ewald sphere.
1 Full reflection.
2 Partial reflection entering the Ewald sphere, extending from previous frame.
3 Partial reflection entering the Ewald sphere, extending to next frame.
4 Partial reflection entering the Ewald sphere, extending to both neighboring frames.
Reflection exiting the Ewald sphere.
11 Full reflection.
12 Partial reflection exiting the Ewald sphere, extending from previous frame.
13 Partial reflection exiting the Ewald sphere, extending to next frame.
14 Partial reflection exiting the Ewald sphere, extending to both neighboring frames.

The value of flag2 is decided based upon the value of flag1 and the integrated intensity and its estimated standard error.
flag2 Description.
0 Default.
1 Predicted by DPS_scale but intensity is non-significant.
2 Not predicted by DPS_scale but intensity is significant.
3 Estimated standard error is equal to, or less than zero.

DPS scale produces 2 reflection output files. The larger of these files contains the scaled reflections in the following format:

The first 3 integers are the reduced Miller indices followed by the centricity. The centricity is - when symmetry operation to obtain the reduced Miller indices from the original Miller indices includes a center of symmetry. The next 3 integers are the original Miller indices followed by the scaled reflection intensity and its esd. Next is the scaled and averaged intensity and its esd followed by the resolution of the reflection.
The smaller of the reflection output files contains the scaled and averaged reflections in the following format:

Again the first 3 integers are the reduced Miller indices followed by the centricity. Only when processing anomalous data are the reflections of opposite centricity kept apart. A ? marks if there is chance of the signs of centricity being mixed. Next is the scaled and averaged reflection intensity followed by its esd and rmsd. The last 2 integers are respectively a flag to mark bad reflections and the number of reflections contributing to the scaled and averaged reflection intensity. Note that when there are no good reflections to contribute to the scaled and averaged reflection intensity, the intensity is set to 0.0000.

Page last updated: January 28, 2000
Authors:
Kai van Beek