Virtual Talk: Overview of the machine and deep learning needs and trends for Single Particle Analysis by Cryo-Electron Microscopy

Carlos Oscar Sorzano
Centro Nacional de Biotecnología (CSIS)

Single Particle Analysis by Cryo-electron Microscopy has established as a key player in Structural Biology as a way to determine the three-dimensional structure of biological macromolecules. Several advances have contributed to this success at different levels: sample preparation, microscope and image acquisition, and image analysis. In this latter step, the number of already existing methods and those appearing is huge. Among them, classical image processing as well as machine and deep learning algorithms have a central role. In this talk we will present an overview of the most active topics and their needs. In particular, there are several pressing needs: robustness to noise, accuracy, speed, automation, and validation. Especially, the one of speed is in contradiction with the other three and automated decisions are not easy to take in absolutely all situations. This is a complicated balance although the field is quickly advancing through it. One of the main tools to have reliable results is the comparison of the estimations of several methods on the same set of images. This is seldom seen in the field at the moment, but these comparisons should be encouraged. Finally, it is important as a field to move into a FAIR (Findable, Accessible, Interoperable and Reusable) data regime (for instance, only 2% of the structures deposited in EMDB for SARS-CoV2 have their corresponding raw data in EMPIAR). In this way scientific transparency will be promoted, and Single Particle Analysis will be based on more solid grounds.

Presentation (PDF File)

Back to Long Programs