Fig. 7
From: Robustness of signal detection in cryo-electron microscopy via a bi-objective-function approach
Effects of different particle-picking templates and starting references in MLE optimization of real-world cryo-EM images of the glucose isomerase complex. The templates used for particle picking were: a Gaussian circle (a, b), one projection view of the glucose isomerase complex (c, d), and one projection view of the HIV-1 Env trimer (e, f). The approximate percentages of false-positive particles assembled in the three cases, estimated through the manual examination of the larger-defocus micrographs in the focal pairs, were 6% (a, b), 4% (c, d) and 11% (e, f). In the MLE optimization step, the unaligned averages of randomly classified particles were used as starting references in panels a, c and e, and a Gaussian circle was used as the starting reference in panels b, d and f. In each panel, the five rows of image series correspond to five particle orientation classes generated by MLE, with the class averages of the milestone iterations (1st, 10th, 50th, 100th, 500th) shown in a row