view article
Figure 1
False discovery rate (FDR) analysis of cryo-EM maps. (a) Left: flowchart of confidence-map generation. The cryo-EM map is converted to p-values and finally FDR-controlled. Right: slice views through a cryo-EM map of 20S proteasome (EMD-6287) depicted at the respective stages of the algorithm (blue boxes) on the left. Note the strong increase in contrast when the sharpened map is converted to the confidence map. (b) Left: estimation of the background noise from windows (red) outside the particle. Right: histograms (top, probability on a linear scale; bottom, probability on a log scale) of the background window together with the probability density function of the estimated Gaussian distribution. (c) Evaluation of the algorithm on a simulated two-dimensional density grid. The upper right quadrant of images in real space (left column) together with the corresponding power spectrum in the Fourier domain (right column) are displayed. A density grid with added normally distributed noise at a signal-to-noise ratio of 1.2 leads to a loss of contrast at high resolution. Confidence maps recapitulate these high-resolution features (arrows), showing that high-resolution signal is detected with high sensitivity. FDR thresholding at 1% recovers a similar binary grid in comparison with 3σ thresholding while minimizing noise contributions and minimizing detected noise (enlarged insets).

IUCrJ
Volume 6| Part 1| January 2019| Pages 18-33
ISSN: 2052-2525
https://doi.org/10.1107/S2052252518014434