In contrast to MIR, where changes in the protein structure are carried out by the inclusion of extra atoms, in MAD changes are induced by changing the scattering strengths of heavy atoms already present. Consequently the observed intensity changes in MAD are smaller than those observed in MIR; typically by a factor of five or more. The implications of this are that intensity differences observed in MAD data owing to anomalous scattering are often of the same order as the experimental errors in the observed diffraction data.
However MAD does offer some advantages over MIR. Firstly, since all the necessary contrast in the diffraction intensities may be obtained using one structure (protein plus heavy atoms) by tuning the X-ray energy only, it is possible to collect all necessary data from the same crystal. This means that non-isomorphism is no longer a problem.
A major source of systematic error in protein diffraction data is absorption of the diffracted X-rays by the crystal sample itself. Small molecule crystallographers often have the option of choosing crystals with regular shapes so that the absorption path length is similar for all diffracted beams. They may also be able to determine the exact crystal shape and apply a correction for crystal absorption. In extreme cases they may even be able to grind their samples into spheres, for example, so that no absorption correction is necessary. These options are not open in protein crystallography. Sample crystals are too sensitive and often too few to allow the ideal shape to be chosen. In addition crystals are often mounted in capillaries and are surrounded by solution. Before calculated absorption corrections can be made it is also necessary to know the exact composition of the crystal which is generally not known early on is a protein diffraction experiment. In MIR the use of three or more different crystals for collecting the different diffraction data sets gives additional complications since each crystal will result in a different absorption path length for each common reflection. MAD allows all data to be collected from the same crystal which can reduce systematic errors. This allows more accurate estimates of the intensity differences to be made.
Collecting more data from one crystal sample inevitably means that radiation damage is more of a problem. It is important to minimise the overall time required for data collection. Recent developments in the cryogenic cooling of protein crystals  have been shown to minimise and in some cases eradicate the radiation damage problem, although this is still a relatively new field of research. The use of high intensity synchrotron radiation reduces overall data collection times and also offers the possibility of using higher X-ray energies where radiation damage is reduced because of the inverse relationship between X-ray energy and absorption coefficient. The time the crystal spends being irradiated should be short  as indeed should the overall experimental time as it is thought that the time dependent nature of free radical migration leads to the continuation of radiation damage effects even when the crystal is not being exposed .
The anomalous signal, observed as differences in intensities, should be measured as accurately as possible. The signal to noise ratio should therefore be maximised. In a MAD experiment this can be done principally in two ways. Firstly, by selecting energies where the anomalous effects are most marked one can increase the anomalous signal with respect to the statistical errors in the data. Secondly one can try and reduce the fractional size of the statistical errors by measuring more photon counts per intensity measurement. A third possibility is to do both of these but the two ideas tend to conflict with one another. Obtaining more accurate data by collecting more counts per intensity measurement necessitates the use of a high intensity source so as not to make data collection time too long. Selecting optimal energies where anomalous scattering factors are large, however, requires the use of X-rays with a good energy resolution comparable to the widths of the anomalous scattering features. Improving energy resolution implies reducing the flux throughput of the X-ray monochromator and will increase data collection times.
The reduction of statistical error by measuring intensities with a larger number of photon counts on average is reliant upon the use of a detector system with a sufficiently large dynamic range. If a number of different images need to be combined to improve statistics this may introduce systematic error resulting from radiation damage or from the scaling together of image.
If the high energy resolution approach is adopted further problems emerge. Positional and angular variation of synchrotron source and monochromator instabilities due to thermal effects all act to make the monochromatic X-ray energy during an experiment an uncertain quantity. The majority of MAD studies have relied upon periodic recalibration of the incident X-ray energy by repeated measurement of fluorescence from the protein crystal or a model compound. This has two disadvantages. First, the X-ray energy during data collection is not known and the user must rely on the source and beam line optics being stable. This is rarely the case. Second, repeated fluorescence measurements from the protein sample will lead to increased radiation damage and may shorten the crystal lifetime.
The problem of X-ray energy stability of synchrotron radiation is one which has not been well covered to date, not only in the field of protein structure analysis by MAD, but in many other areas of synchrotron radiation which utilise anomalous scattering effects. There is a genuine need for a generally applicable method of X-ray energy calibration and stabilisation which can be used during a diffraction experiment and which does not rely on repeated periodic interruption of the data collection procedure and additional irradiation of the crystal sample.
This thesis addresses the practical difficulties associated with the measurement of X-ray diffraction data at X-ray energies where the anomalous scattering factors of atoms within the crystal structure are highly sensitive to shifts in the X-ray energy. i.e. around the characteristic absorption edges of atoms.