THE ESSENTIAL PRELIMINARY TO MAD PHASING; finding the positions of the Anomalous scatterers.

J. Yao, E. Dodson

Department of Chemistry, University of York, Heslington Y01 5DD, UK yao@yorvic.york.ac.uk E.Dodson@yorvic.york.ac.uk

Keywords: Locating Anomalous scatterers, Direct Methods.

The exploitation of anomalous dispersion edges to determine experimental phases has been shown to be a powerful tool for solving macromolecules. Suitable heavy atoms whose edges are at accessible wave lengths using synchroton resources are Se and Br which can be substituted chemically into proteins and DNA. Se can substitute the Sulphur atom in methanine residues if the protein is produced under suitable conditions. However for larger proteins there may be 20 - 100 Se sites, and it is difficult to find these using classic Patterson techniques.

The estimates of the anomalous scatterers contribution(FA) must be deduced measured at different wave lengths, which are due to differences in the f' for these atoms; and from the anomalous differences between Friedel pairs, due to the f" values.. The formal mathematics is well-established, given knowledge of the appropriate f' and f" for the wavelength. [1]

It has been shown that if providing suitable estimates of these FA are available classic patterson search or direct methods can find their positions. Patterson search methods usually fail once the number of sites is greater than 10, but direct methods can succeed to find many sites from limited and incomplete data. This has been demonstrated using calculated structure factors from the 52 Se sites with temperature factors ranging from B=11 to B=53 in the structure, 1ECF. [2] The calculation was performed using data to 3 A in the direct methods program RANTAN [3]. The default figure of merits clearly indicated the correct solution. However in a realistic experiment many of the FAs are not observable, since they are associated with weak, and hence less reliable, observed structure factors. There is very little correlation between weak Fobs, and weak FAs. To test the method more rigorously, a randomly selected 50% of these calculated structure factors were excluded from the direct methods procedure.

Again in this case the correct result was easy to recognise. Again both methods for determining the scatterer's positions are extremely sensitive to large incorrectly estimated outliers, and can only work when most of these are detected and weeded out.

It can be shown that

a) careful data collection and processing can help to detect outliers.

b) Checking the ratio of differences between different wave length pairs helps estimate the f' and f" values. In particular normal probability plots [4] for the different data pairs are a useful tool for verifying the values of f' and f".

c) Revising these difference to keep the expected ratios constant improves the estimate of FA and can reject more outllers automatically. [5,6] All these steps can thus improve the performance of the direct methods.

1. Hendrickson,W.A. (1991). Science 254, 51-58.
2. Muchmore_CRA, Krahn_JM, Kim_JH, Zalkin_H, Smith_JL (1198) Vol 7 No 1 39-51
3. Yao Jia-xing (1981). Acta Cryst. A37, 642­644.
4. Smith,D and Howell,L. J. Appl. Cryst. (1992) 25 81-86)
5. Woolfson,M.M., Yao Jia-xing & Fan Hai-fu. (1997). Acta Cryst. D53, 673­682.
6. Fan Hai-fu, Woolfson,M.M. & Yao Jia-xing, (1993). Proc. R. Soc. Lond. A 442, 13-32