Stefan and Magdalena are correct about the importance of Friedel pairs, as the differences between the pairs basically provides the anomalous information. There were a couple of presentations at the last international CCP4 meeting in the UK where the main claim was that alignment along the appropriate axis using a Kappa goniometer often worked much better than the inverse beam mode. By aligning the crystal with the beam in such a way that one of the cell axes (usually the longest) lies perpendicular to the beam, you obtain diffraction images that have bilateral symmetry and where both parts of a Friedel pair are visible on the same image. This way the differences between each pair are easily measured and compared and, because they are measured at the same time, the differences cannot be due to X-ray damage, scaling or anisotropy that result when the two parts of a Friedel pair are measured from different images taken from different positions on the crystal at different times.

At the Diamond Light Source they are now implementing these goniometers on all the macromolecular beam lines, you simply take 2 to 4 test images between 0o and 90o and the software automatically spits out the theta and kappa angles you need to orient the goniometer. After you collect the data set you should then collect a second data set where you misalign all the crystal axes. This is useful since the first data set will have no data for the aligned axis and it will help determine the correct space group if there are systematic absences along that plane. By integrating both data sets and then merging and scaling them together you will obtain the highest completeness possible

This method is useful for anomalous diffraction and also for native crystals where one of the axes is so large that all the reflections along one plane are so closely set in the diffraction image that they merge with one another and with reflections from the other planes. Closely set reflections are discarded by the integration software, so you lose a lot of data and get bad completeness in these cases. However, if you align the long axis, you won't see the closely set reflections at all and you can collect 100% of the data for the other two axes. You then misalign the crystal or align it along one of the other two axes and collect as much data as you can from the plane of the long axis. Many of the reflections from the long axis will still be discarded, but at this point it doesn't matter if they interfere with reflections from the other axes as they have already been collected in the previous data set. Again, merge the data sets for the best completeness possible.

Dear Heng,

The desired resolution for anomalous phasing is often lower than you think. Obviously, a high overall resolution is always desirable, but for anomalous phasing you also need to look at the anomalous resolution (they are not the same). You can potentially solve good data (no twinning, pseudo-translational symmetry, etc...) even at an anomalous resolution as low as 5 to 6 Angstroms, though in practice you generally need at least 3 to 4 Angstroms of anomalous resolution.

However, there are a number of factors that influence the result: If your crystals are stable, decay very slowly in the X-ray beam and you have a high symmetry space group, you can attempt MAD phasing by collecting 3 short data sets (e.g. 90o each for an orthorhombic space group) at 3 different wavelengths (these will depend on the particular heavy metal in the soak) from a single crystal. If your crystals decay rapidly and you have a low symmetry space group, you should try SAD phasing. In this case it's best to collect a lot of data (360 to 720o) close to the peak wavelength with high attenuation, fast exposure times and slightly larger oscillations than usual (1 to 2o). If your crystals are isomorphous (different crystals have almost identical unit cell parameters) then you can merge the data from several crystals to improve your SAD phasing as the anomalous resolution often increases with high multiplicity (the overall resolution does not).

Heavy metals with large anomalous signals (e.g U, Hg, Au, Pt) are easier to find and solve than smaller atoms with less scattering power. Small numbers of heavy metal atoms are easier to locate than large numbers, particularly if the atoms are small in size and can lie close to each other, as is the case with Se in selenomethionine substituted protein or with halide soaks (Br and I can give several dozen and even hundreds of sites).

The heavy metal atoms will be damaged first at the wavelength at which they produce anomalous diffraction, so the best anomalous resolution will be at the start of the data even if the overall resolution doesn't seem to decrease, so check if you get better data by discarding the images to the end of the collection. This is also why combining the first block of images from several isomorphous crystals in SAD gives better results in many cases.

You can try a multitude of different things to see what works best if you have many crystals. However, if you only have a few, try to obtain the best high resolution you can with one crystal and then try high attenuation and fast exposures to collect data to 3 Angstrom with another crystal. Then compare the extend and quality of the anomalous signal of both data sets to see which strategy works best before you collect data from further crystals.

I hope this helped.

Hi Antonio, 

Thank you very much for you brilliant explanation. Can you tell me more about anomalous resolution? I am not quite understand what it is. Do you have any recommended sources for me to read? 

Anomalous resolution limit are the highest resolved reflections that still yield a significant anomalous difference.  While there are more higher resolution reflections on the detector, those won't help with substructure determination and initial phasing. However, they are useful in later steps, if the substructure is correct.

Another thing to consider is anomalous completeness. If you set up data collection for anomalous data you want to use a strategy that allows you to observe all Friedel pairs rather than all reflections. Mosflm and other data processing tools have an option for that. You just need to collect 2-3 frames, process the data and let the software find strategies to collect most Friedel pairs in the shortest time. 

You want precisely measured anomalous differences. The resolution of the anomalous signal is not that important as long as the data is good!

I completely agree with Antonio and Stefan. Stefan mentioned anomalous data completeness. We usually collect data in the inverse beam mode: we collect some degrees of data and then we rotate the crystal by 180 degrees and collect the same amount of data (and so forth). In that way you can make sure that we have recorded all the anmalous mates, before radiation damage kicks in. The important thing is that your detector is centered. Most of the synchrotron beamlines have an automatic script for this kind of data collection. Good luck, Magdalena

Thanks Stefan and Megdalena for your valuable suggestions!

Stefan and Magdalena are correct about the importance of Friedel pairs, as the differences between the pairs basically provides the anomalous information. There were a couple of presentations at the last international CCP4 meeting in the UK where the main claim was that alignment along the appropriate axis using a Kappa goniometer often worked much better than the inverse beam mode. By aligning the crystal with the beam in such a way that one of the cell axes (usually the longest) lies perpendicular to the beam, you obtain diffraction images that have bilateral symmetry and where both parts of a Friedel pair are visible on the same image. This way the differences between each pair are easily measured and compared and, because they are measured at the same time, the differences cannot be due to X-ray damage, scaling or anisotropy that result when the two parts of a Friedel pair are measured from different images taken from different positions on the crystal at different times.

At the Diamond Light Source they are now implementing these goniometers on all the macromolecular beam lines, you simply take 2 to 4 test images between 0o and 90o and the software automatically spits out the theta and kappa angles you need to orient the goniometer. After you collect the data set you should then collect a second data set where you misalign all the crystal axes. This is useful since the first data set will have no data for the aligned axis and it will help determine the correct space group if there are systematic absences along that plane. By integrating both data sets and then merging and scaling them together you will obtain the highest completeness possible

This method is useful for anomalous diffraction and also for native crystals where one of the axes is so large that all the reflections along one plane are so closely set in the diffraction image that they merge with one another and with reflections from the other planes. Closely set reflections are discarded by the integration software, so you lose a lot of data and get bad completeness in these cases. However, if you align the long axis, you won't see the closely set reflections at all and you can collect 100% of the data for the other two axes. You then misalign the crystal or align it along one of the other two axes and collect as much data as you can from the plane of the long axis. Many of the reflections from the long axis will still be discarded, but at this point it doesn't matter if they interfere with reflections from the other axes as they have already been collected in the previous data set. Again, merge the data sets for the best completeness possible.