Insensitivity to outliers

This example shows results from an anomalous Patterson function calculation using data collected from a horse heart myoglobin crystal. The data were collected with CuK radiation and the anomalous signal comes from the iron atom of heme (with f_{Fe, CuK}'' = 3.2e^-). To make the example more realistic we only used data between 20 and 3Å resolution, and we simulated the presence of outliers in the data by multiplying the amplitude ( F_ano) and standard deviation ( (F_ano)) of three randomly chosen strong reflections by a factor of 3.0.

A comparison of the Harker sections (v = 1/2) from the conventional and GraphEnt maps (shown in the figure that follows) is rather striking : The presence of outliers in the data has completely wiped-out the signal from the conventional map (left-hand-side panel), leaving behind the only too familiar to macromolecular crystallographers chequer-board appearance. In sharp contrast, the GraphEnt map resembles more a map calculated with hypothetical error-free data than an anomalous Patterson function calculated with real data (both maps are contoured with the dashed contour at the mean, and then every 0.5 rmsd of the whole map).

Figure: Insensitivity to outliers