A Two-Stage Dimension Reduction Method for Cryo-EM Image Processing

With the recent breakthrough in the camera together with microscopy automation and advancement of algorithms, cryo-EM has become a mainstream technique to solve structures of macromolecules at near atomic resolution. As a result, single particle cryo-EM has awarded Nobel Prize in chemistry in 2017. A renowned feature of cryo-EM image data is its extremely low signal-to-noise ratios (SNRs) due to usage of low-dosed electron beams for imaging to avoid radiation damages on biological structures. In addition, single particle cryo-EM data has acquired another character in the wake of resolution revolution, which makes its processing even more challenging. Owing to the current advances of electron microscope detector and automation, the size of a cryo-EM particle image is often larger than a hundred in each direction and the total number of particles exceeds hundreds of thousands such that processing cryo-EM images is thereby transformed into a high-dimension problem.