The signal-to-noise ratio in SAD experiments

De novo single-wavelength anomalous dispersion (SAD) determination of macromolecular structures requires accurate measurement of anomalous signal from naturally occurring (S, P, Ca, etc.) or incor- porated (Se, Hg, etc.) anomalous scatterers. The quality and the level of the anomalous signal in SAD datasets can be assessed using, as an indicator, the average anomalous signal-to-noise level ⟨|Ianom|/σ (I)⟩ based on intensities or ⟨|Fanom|/σ (F)⟩ based on amplitudes. The value for the average anomalous signal-to-noise for a successful SAD experiment is a matter of debate given its high vari- ability from one SAD experiment to another. We present a mathemat- ical model relating the ⟨|Ianom|/σ(I)⟩ to the overall final ⟨I/σ(I)⟩, necessary for successful SAD phasing. By statistical analysis of the data from 115 successful native SAD experiments, as available in the PDB, we show that the experimental values of ⟨|Ianom|/σ(I)⟩ fol- low a Gamma distribution with an average value of 1.2±0.3. This mathematical model allows a-priori prediction of the overall ⟨I/σ (I)⟩, necessary for successful SAD phasing given the description of the sample expressed. The results reported here are of general appli- cability to any phasing experiment involving the measurement of anomalous signal from any anomalous scatterer or X-ray source, including X-FEL data