Fig. 4
From: Robustness of signal detection in cryo-electron microscopy via a bi-objective-function approach
The correlation-peak ranking plots and differentiation of true-positive and false-positive particles in FLC-based automated particle picking. The correlation-peak ranking plots corresponding to different SNRs, obtained using three different particle-picking templates: (a) a Gaussian circle, (b) one projection view of the influenza virus HA trimer, and (c) one projection view of the HIV-1 Env trimer. The particle-picking templates are shown in the insets. All plots are from the noisy particle micrographs derived from the same simulated noiseless micrograph of the influenza virus HA trimer. Note that the position of the drop-off in the correlation peak values corresponds to 323, which was the number of actual influenza virus HA trimers in the simulated micrographs. (d) Rate of false positivity in particle picking. The plots of false positive fraction against SNR in particle picking using the three different templates are shown, indicating that the specificity of FLC particle picking is highly dependent on the SNR, and is also affected to a lesser extent by the choice of the 2D template. Below a critical SNR range (0.002–0.005), the percentage of false positives rises considerably