Single-particle cryo-EM reconstructs 3D molecular structures from noisy projection images, and the standard mathematical model assumes each image contains exactly one particle centered at the origin. In practice, many experimental datasets include micrographs with a central particle plus a few crowding particles, and this deviation can bias estimates of low-order image statistics such as image moments.
In a new arXiv preprint, Andy Zhang, Oscar Mickelin, and Amit Singer propose a method to estimate image moments while explicitly correcting for the effect of crowding particles. The approach can be run directly from a set of cryo-EM micrographs with picked particles, avoiding the need to discard or relabel crowded images.
The authors demonstrate the method's efficacy on a wide array of synthetic data, showing that the correction improves moment estimation under simulated crowding. Because the paper reports only synthetic experiments, its performance on real experimental datasets remains an open question.