AVRGAMPHS
By Vinzenz Unger and Anchi Cheng
AVRGAMPHS is used to average projection data from several images. Because of a scaling error in older versions, version 1.5 or higher should be used. The in- and outputfile are set by FORT1 and FORT2 respectively. The program expects a list of data written by ORIGTILTC/D if the flag on the first input card is set to “T”. Otherwise a format like that produced by ORIGTILTB is assumed. The image data can contain “non-projection” data points that are derived from images of tilted crystals. However, an appropriate setting of the z*-cutoff values is important to ensure that only projection data contribute to the average. Since the exact specimen thickness will not be known in most cases it is recommended to use a starting value of [1/2x estimated specimen thickness] for the averaging. This setting is not to be confused with choosing an appropriate setting for the width of the real space envelope in LATLINED, which should be as close to the real specimen thickness as possible.
The program will return a list of (H K L) and their respective CTF-scaled average amplitude, an amplitude weighted average phase and a “figure of merit” (FOM) that reflects the amplitude weighted, average phase error associated with each reflection. In addition the program provides useful diagnostic outputs and overall statistics that are useful to evaluate the result of the averaging procedure. As mentioned, the program calculates a CTF-scaled average amplitude. Correction of the amplitudes (as well as the phases) for the contrast transfer function is important if no electron diffraction data are available. A number of different scaling procedures are employed by AVRGAMPHS to deal with different types of input data.
If multiple measurements are provided and at least one of the data points has an IQ of 3 or better, amplitudes will be corrected by division by the absolute value of the CTF (found in the last column of the ORIGTILTD output) before combination. To avoid problems for measurements that are close to a CTF-zero, a maximum 5-fold scaling is applied to data that have absolute CTF-values of ≤ 0.200. This amount of scaling is arbitrary but has been shown to give satisfactory results. It is not recommended to change this unless you have good reasons and any changes should be stated upon publication. If the average scaling factor for measurement of a particular reflection exceeds 2.5 a “warning message” will appear and the explicit input data for this reflection will be listed. This “warning” is merely to notify the operator of a larger scaling factor rather than a reason for concern.
A different strategy of amplitude scaling is employed if none of multiple input data points has an IQ of 3 or better. In this case a maximum 2-fold scaling will be applied. Regardless of the signal-to-noise ratio (IQ - value) a maximum 2-fold scaling is also employed if only a single measurement is provided. Both measures limit the amplification of data that are either weak or noisy or do not allow a proper assessment of the satistical significance due to the lack of redundancy.
Lastly, it should be emphasized that AVRGAMPHS produces an amplitude weighted vector average for the phase value of each structure factor. If the specimen has symmetry that constrain certain phases to be real (i.e. 0 or 180˚) or absent, appropriate adjustment needs to made either manually or by running a “jiffy-program” like FOMSTATS (see ftp-side). With regard to the experimental phase value, rounding to the closer theoretical phase - where applicable - constitutes the introduction of an additional “phase error” to the experimental data and, hence, the FOM value calculated by AVRGAMPHS should be corrected for the deviation from 0/180˚ to account for the degree of disagreement between the experimental and symmetry constrained phase value. In contrast to X-ray crystallography, the final phase errors are the best and most objective criterion to evaluate the quality of the data. For projection data a representation of the phase errors of each individual reflection can be produced by the program PLOTALL. Including a detailed account of the phase statistics in publications or at least for the reviewer is strongly recommended as they easily allow an unbiased observer to make an independent judgment on the reliability of any density maps shown.