Figure 1
Eigenvalues of the PCA matrix for five series of datasets: (a) thaumatin, (b) Au-derivatized elastase, (c) Pt-derivatized elastase, (d) Lon protease, (e) thermolysin. Centric and non-centric reflections were used in the data analysis with the Friedel pairs merged. The native signal represents 100% and corresponds to the first eigenvalue, which was removed from the figures to show clearly the contributions from other components. The orange bars represent the second eigenvalue, the purple bars represent the third eigenvalue, and the blue bars represent the remaining eigenvalues. The insert shows the same plot after removing the second eigenvalue, so it can be seen how the third eigenvalue compares with the subsequent ones. The plots show that in the case of thaumatin and Lon protease there are two significant (above the noise level) components of radiation-induced specific changes, whereas for thermolysin and elastase the second component of radiation-induced specific changes has borderline statistical significance. The noise components were defined by analyzing ratios between consecutive eigenvalues, i.e. 2:3, 3:4 and so on. If these ratios are only slightly above 1, these eigenvalues are interpreted as noise components. The components preceding such a group have a combination of noise and signals. If the preceding component has a ratio to the next one that is less than 2.0, the signal in it is interpreted to be smaller than the noise; the signal can still be statistically significant, but is not very informative.  [article HTML]

© International Union of Crystallography 2013