Table 3.

A comparison of the different variables that affect structural studies undertaken by X-ray crystallography and single particle Cryo-EM.

	X-ray Crystallography	Cryo-EM
Protein size range	Average size of solved structures is ~100 kDa [7].	Typically above 100 kDa. Volta Phase plates have been used to boost signal-to-noise for smaller proteins or binders such as Fabs and megabodies can be used to increase particle size [163,217].
Sample heterogeneity	Usually a homogeneous sample is required [229].	Can tolerate some sample heterogeneity but homogeneous samples lead more quickly to higher resolution structures [216].
Sample concentration	Large quantities of pure protein [230]. Typically 100 to 200 µL at 5 to 40 mg mL−1.	Small quantities of pure protein [230]. Less than 10 to 100 µL at 0.5 to 5 mg mL−1.
Sample preparation	Relies on obtaining diffracting MP crystals which are difficult to obtain. MP must be removed from its native environment. Crystals grown in crystallisation trays, mounted in a loop and cryo-cooled in liquid nitrogen [171].	MP blotted on to EM grids and vitrified in liquid ethane [110,231]. Single particle analysis can be carried out on proteoliposomes providing a more native environment [232].
Screening throughput	High. Typically in 96 well plates, allowing 100s of conditions to be sampled simultaneously [171].	Low. Each condition to be screened must be imaged individually. Negative stain can be used to narrow screening conditions [216].
Collection method	X-ray diffraction of protein in crystalline lattice, typically using a synchrotron source [171]. Microcrystal electron diffraction is an area of increasing interest [233].	Electron imaging in conjunction with a direct electron detector. Energy filters and phase plates may be helpful [234].
Collection throughput	High. Typically, 15–30 crystals per hour [7].	Low. Time taken several orders of magnitude behind X-ray crystallography [235].
Data Analysis	Quick and highly automated. Complete datasets can be collected in seconds. Many synchrotrons have automated processing pipelines integrated into the data collection process [7]. Well established software suites such as CCP4i2 to aid the crystallographer [236].	Slow. Reconstructions from 1000s of single images can take many days. Processing pipeline can be automated. Software packages to analyse data less established but constantly improving. Examples include RELION and cryoSPARC [237,238].
Structure-based drug design	Routine, high resolution and high throughput. Well established for GPCRs [13,239].	Currently lacks reproducibility, quality and throughput. Ideally requires protein structures at a resolution of less than 3 Å [240,241].
MP conformational flexibility	Each crystal form relates to a single rigid MP conformation.	MP can be in different conformations, which can be identified during processing (but also impede processing) [216].
Ion identification	Generally straightforward depending on resolution. Long-wavelength beamlines enables sodium ion to be distinguished from a potassium ions [228].	Difficult to identify some anions ions in maps due to negative scattering factors [242]. Electrostatic potential maps may help to overcome this [243].
Resolution	Typical range between 1.5 Å and 3.5 Å. For MPs crystallised in LCP Sub 2.5 Å are common. Highest resolution structure currently a yeast aquaporin at 0.88 Å, PDB: 3ZOJ [244].	Typically, 2.5–4 Å are common including some smaller membrane proteins [217]. Highest resolution structure currently the β3 GABAA receptor homopentamer at 1.7 Å, PDB: 7A5V [219]. EM density maps can identify protein and ion charge states [242].