From its beginning in the late 1970’s, cryo-electron microscopy (cryo-EM) has been steadily refined for structure determination in the biological sciences. Now approaching 3Å resolution, cryo-EM was named “Method of the Year” by Nature Methods [1]. Advances on several fronts were required: sample preparation, electron detection, methodology, and data processing algorithms.
Signal-to-noise ratio (S/N) has driven many if not most of these developments. A look at the micrographs in [2] is quite revealing in this regard. Even with electron doses which e-beam lithographers would shun in the certain knowledge that their photoresist would char (30 e–/Å2 ≈ 48000 μC/cm2), image contrast is quite poor. Direct electron detectors (a.k.a. thinned substrate CMOS) have helped significantly. (See [1], pp. 19-22 for some history.)
The electron dose damages the target molecules, encouraging the use of movie-mode data collection. Data processing recombines frames after removing image motion. It is possible to use the integrated intensity for low-resolution steps in the data processing chain (particle picking and averaging), while selecting only early frames showing less damage for the final high resolution determination. (There’s an obvious trade-off here.)
Phase contrast imaging should vastly improve the limitations due to S/N, although this is not evidenced by the first attempt [3].
Cryo-EM can be thought of as a method of tomography wherein thousands of identical target molecules are imaged in random orientations, instead of one target being imaged in multiple, chosen orientations. The first data processing step is to choose the subfields containing relevant images. Given the poor contrast, this step is fraught with uncertainty and demands some type of validation methodology. (See ref [4], for example.) It is entirely possible to extract a picture of Albert Einstein from random noise if one “looks” hard enough [5].
Cryo-EM is compute intensive. We’ll return to this point in a later post. Until then, I will refer to a recent review listing several software packages [6], and another review discussing recent successes and some limitations [7].
References:
[1] “Special Feature” papers in Nature Methods 13(1)2016. DOI: 10.1038/nmeth.3730, 3698, 3700, 3694, 3495
[2] van Heel, et al., “Single-particle electron cryo-microscopy: towards atomic resolution”, Quarterly Reviews of Biophysics 33, 4 (2000), 307–369. DOI: 10.1017/S0033583500003644
[3] Danev and Baumeister, “Cryo-EM single particle analysis with the Volta phase plate”, eLife 2016;5:e13046. DOI: 10.7554/eLife.13046
[4] Falkner and Schröder, “Cross-validation in cryo-EM–based structural modeling”, PNAS 110 (2013) 8930-5. DOI: 10.1073/pnas.1119041110
[5] Henderson, “Avoiding the pitfalls of single particle cryo-electron microscopy: Einstein from noise”, PNAS 110 (2013) 18037-41. DOI: 10.1073/pnas.1314449110
[6] Elmlund and Elmlund, “Cryogenic Electron Microscopy and Single-Particle Analysis”, Annu. Rev. Biochem. 2015. 84:20.1–20.19. DOI: 10.1146/annurev-biochem-060614-034226
[7] Bai, et al., “How cryo-EM is revolutionizing structural biology”, Trends in Biochemical Sciences 40 (2015) 49. DOI: 10.1016/j.tibs.2014.10.005