CCP4i: Graphical User Interface
	Tutorial 2

Molecular replacement

AMoRe - more involved examples

The correlation coefficients and Rfactors as they came out of the simple AMoRe example are not particularly promising. It is, therefore, wise to play around with some of the parameters.

Do not worry about the jumping of windows. Whenever only a shape of a window comes up, position it on the screen where you want it (as before).

Varying the resolution limits

Using all available data is not necessarily a good thing in molecular replacement calculations. The highest resolution data are involved with detail, potentially confusing the overall shape of the search model. Besides, the calculations take longer with higher resolution. The lowest resolution data contain information on the solvent rather than the protein, and can thus be taken out of the molecular replacement calculations. It is thought that cutting off the resolution between 8 and 12Å at the low end, and between 3 and 5Å at the high end, is a good idea.

1. 12-3Å

   To run auto-AMoRe with the resolution cut off at 12 and 3Å, start as before. You might want to change the Job title to reflect the changes made to the input. The input files can remain the same. It does no harm to have the model and crystal 'tables' ready with all data. The Interface will skip the SORTFUN and TABFUN stages.
   In the 'Key Parameters' folder, change the Resolution range minimum and maximum to
   12 and 3 respectively.
   The Interface adds decimal points to these numbers.

   Because the model table was prepared without rotation, the output from the calculations will be readily interpretable in terms of expected twofold rotations, and also comparable to the output from the simple example.

   Select Run > Run Now
   at the bottom left of the Task Window to set AMoRe running.

2. 10-3Å

   Rerun AMoRe as above, with resolution limits 10 and 3Å

3. 10-4Å

   Rerun AMoRe as above, with resolution limits 10 and 4Å

4. 8-3Å

   Rerun AMoRe as above, with resolution limits 8 and 3Å

5. 8-4Å

   Rerun AMoRe as above, with resolution limits 8 and 4Å

6. 6-4Å

   It is sometimes tempting to narrow down the resolution range even more. Rerun AMoRe as above, with resolution limits 6 and 4Å

Evaluating the input and output

From the Command Scripts it can be seen that the Interface indeed skips the SORTFUN and TABFUN stages, simply recycling their output.

1. View Output solution files for example 1 (12-3Å)

   ..._rot_...mr

   shows two peaks higher than half the maximum. They are equivalent, and the same as in the simple example, with similar solution criteria (24.6-53.2-37.9-28.3 in this search, and 25.4-56.0-24.6-24.9 before). CC_I is markedly different and may be a better criterium to distinguish correct from noise in this case.

   ..._tran_...mr

   shows a much clearer picture than the translation function output of the simple example. The first translations for the first two rotation function peaks are more than twice as high as the highest noise peak, which is why no noise peaks are included in the list at all. In order to include more peaks, you would have to lower the threshold fraction of maximum peak height in the Translation Function Parameters folder and rerun AMoRe. This would show noise peaks with correlation coefficients no higher than 48 and Rfactors well in the 40s, while the correct peaks have values of around 70-75% and 30-35%. N.B. the peak height and the correlation coefficients have no directly proportional relationship, which is why a noise peak with a cc of 48% may be lower than half the peak height of a correct peak with a cc of 75%.

   ..._fit_...mr

   shows an even more convincing picture than the translation function. Peaks with correlation coefficients in the 80s and Rfactors below 30% can be believed directly.

2. Table with output for other examples
   • In order to obtain values for 'minor' peaks, the threshold fraction of maximum peak height may have to be reset to a lower value, both in the Rotation Function and the Translation Function Parameters folders.
   • Within the TRAFUN part of the table, the second translations have been included for the major peaks, in order to show the height of the noise within the peak relative to the noise from other rotation function peaks.
   • In order to have some idea of the relative order of noise peaks, not only the third rotation function peak is included, but also the peak that ends up as the third highest after the translation function.
   • CC_F has been used to sort the peak lists. Please note that CC_F is not a directly proportional consequence of the peak height. The peak height is a measure of the Patterson correlation (CC_P).

resolution range	ROTFUN					TRAFUN					FITFUN
	peak nr	CC_F	RF_F	CC_I	CC_P	peak nr	PHCª	CC_F	RF_F	CC_I	peak nr	CC_F	RF_F	CC_I
10-3	1	22.5	53.0	35.0	29.2	2	1	70.9	34.1	69.8	2	83.2	26.8	83.2
							2	42.3	46.9	44.2
	2	23.4	52.8	36.3	29.2	1	1	74.5	32.0	72.8	1
							2	44.1	46.0	45.7
	3	8.1	56.9	13.6	12.7	7	5	13.8	56.5	14.1	7	24.1	55.0	24.0
	4	9.8	56.5	14.5	12.2	3	1	21.1	54.2	19.6	3	30.1	53.0	28.0
10-4	1	15.6	53.3	29.3	29.4	2	1	71.7	32.2	70.2	2	81.9	26.1	82.0
							3	42.8	45.0	44.4
	2	15.9	53.2	29.8	29.4	1	1	72.9	31.5	71.4	1
							2	42.2	45.1	43.9
	3	1.1	57.0	4.1	14.7	5	2	11.3	55.8	12.2	5	17.6	53.0	19.9
	7	2.2	56.8	5.1	13.0	3	1	14.0	54.8	13.4	3	19.1	53.2	19.7
8-3	1	24.0	52.3	39.9	29.4	1	1	76.3	30.9	75.4	1	83.4	26.6	83.9
							2	45.4	45.3	47.3
	2	23.1	52.6	38.8	29.4	2	1	72.9	33.0	72.6	2
							2	45.4	45.3	47.3
	3	7.8	56.8	14.0	11.9	7	3	13.7	55.9	14.3	7	22.8	54.8	22.4
	4	9.6	56.3	14.7	11.9	3	1	21.2	53.9	20.5	3	29.2	52.8	27.8
8-4	1	16.2	52.9	34.4	29.7	2	1	74.6	30.6	74.1	2	82.4	25.8	83.3
							2	44.9	44.1	47.5
	2	16.5	52.8	34.9	29.7	1	1	75.6	30.1	75.1	1
							2	45.4	44.0	47.9
	3	2.3	56.7	5.3	14.1	6	1	12.3	55.0	12.3	6	14.1	54.2	14.9
	17	2.6	56.8	4.9	12.0	3	4	14.1	55.4	13.4	3	18.0	54.1	17.0
6-4	1	17.6	52.1	38.2	28.0	2	1	79.0	27.6	80.3	2	84.5	24.0	85.8
							2	47.9	42.2	51.3
	2	17.6	52.1	38.3	28.0	1	1	79.2	27.5	80.4	1
							2	48.0	42.2	51.4
	3	4.2	55.8	8.3	15.7	8	8	13.6	54.9	15.6	8	16.0	53.2	17.5
	4	2.4	56.0	6.6	15.7	3	1	15.4	54.0	15.4	3	17.3	53.0	16.1
ª PHC=PkHtCounter

Conclusions

• There is usually an optimum in the resolution range used for molecular replacement calculations. It can be seen from this collection of examples (including the simple one in the previous tutorial) that this is not really the case for RNAse SA. The correlation coefficients of the correct peaks vary a bit, as well as the number of peaks higher than half the maximum (i.e. the amount of potential noise peaks creeping in).
• Sorting peaks on CC_F is not always optimal. Also, especially in ROTFUN, RF_F is not very useful as discriminator of validity of a peak.
• The rotation function calculations are not very exact, which results in the solution criteria for equivalent peaks varying slightly. This gets rectified during the fitting function. The order of the peaks also varies markedly through the stages. What is the third highest peak in ROTFUN, is not usually the third highest in TRAFUN and FITFUN. However, the correct solutions are always recognisable in the end.

More information

For more information on the AMoRe program, see AMoRe.

In the MRTutorial(Bath) there is also a very brief mention of the influence of resolution limits. The case described there requires quite a different resolution range for optimal results.

The Interface documentation on features from this tutorial can be found at: