journal menu
electron diffraction
 |
STRUCTURAL CHEMISTRY |
access
Standards for MicroED
aDepartment of Chemistry, Umeå University, 901 87 Umeå, Sweden, bChemical Engineering, School for Engineering of Matter, Transport and Energy, Arizona
State University, Tempe, AZ, USA, cBiodesign Center for Applied Structural Discovery, Biodesign Institute, Arizona State
University, Tempe, AZ, USA, dMolecular Structure Facility, 149 Stepan Chemistry Hall, Department of Chemistry and
Biochemistry, University of Notre Dame, Notre Dame, IN 46556, USA, eHoward Hughes Medical Institute, University of California, Los Angeles, Los Angeles
90095, USA, fDepartment of Physiology, University of California, Los Angeles, Los Angeles 90095,
USA, and gDepartment of Biological Chemistry, University of California, Los Angeles, Los Angeles
90095, USA
*Correspondence e-mail: [email protected]
This article is part of the collection Advances in electron diffraction for structural characterization.
As the electron diffraction technique MicroED gains momentum and is increasingly embraced by researchers in both academia and industry, we have the opportunity to familiarize ourselves with the characteristics of MicroED data and results. The number of refined structures and their associated data is steadily growing and becoming more accessible to the scientific community, offering valuable insights into the significance and quality of MicroED-derived structures. Additionally, the growing body of experience is helping to identify best practices for the technique. In this summary, we highlight key lessons learned from these data and propose gold standards for the community to consider adopting.
Keywords: MicroED; electron diffraction; quality indicator; best practices; cryoEM; Gold standards.
1. Introduction
Over 100 years of crystallography started when Max von Laue with colleagues demonstrated
that matter could diffract X-rays using a copper sulfate crystal in 1912. The beginnings
of the development of an electron microscope started decades later. Louis de Broglie
proposed in 1924 that electrons also have wave properties, such as wavelength and
frequency, and this was quickly followed by confirmation of electron diffraction
by Clinton Davisson and Lester Germer. An early application of the electron microscope
was presented by Max Knoll and Ernst Ruska in 1931, and electron crystallography could
be considered as starting with the first analysis using electron diffraction (ED) presented in 1949 by Pinsker and Vainshtein
(Vainshtein & Pinsker, 1949![[Vainshtein, B. K. & Pinsker, Z. G. (1949). J. Phys. Chem. SSSR, 23, 1058-1067.]](https://journals.iucr.org/logos/arrows/c_arr.gif "Vainshtein, B. K. & Pinsker, Z. G. (1949). J. Phys. Chem. SSSR, 23, 1058-1067."){kind=small}
). However, after this initial progress, the field of electron crystallography was
quickly set back by the difficulties in interpreting the diffraction patterns and
correlating them to the structure of the molecules in the crystals. Cowley soon found
that the two-beam theory was not sufficient to explain the electron diffraction patterns,
and a dynamic scattering theory was required for electron diffraction data for heavier
atoms (Cowley, 1953), such as lead carbonate (Cowley, 1956). Later, it was argued that simulations of showed that phasing could be realistically possible (Dorset et al., 1979). Cowley also realized that by using a smaller beam and thinner samples, dynamical
scattering could be minimized. Early approaches to macromolecular electron crystallography
typically involved still diffraction images taken at several fixed tilt angles (typically
0, 20, 45, and 60°). For radiation-hardy materials, rather than proteins, it was found
that solving a structure ab initio was possible with precession electron diffraction, rather than still images, which
to some extent averaged out effects of dynamical scattering (Vincent & Midgley, 1994), and later this application for materials structure matured with additional approaches
(Andrusenko et al., 2015; Oleynikov et al., 2007). However, these techniques require precise alignment of the crystal axis on the
rotation axes and in the electron beam. This requires a significant amount of expertise,
specialized equipment, and data processing procedures, and can be difficult for
samples with larger unit cells such as macromolecular crystals. Importantly, the method
requires a very high exposure, limiting its usefulness for biological samples like
proteins that cannot withstand the high radiation, making precession unsuitable for
these samples.
In parallel with the work in materials science, electron crystallography for biological
samples was successful for the structure solution of 2D crystals, and provided some
of the earliest structural information on membrane protein structure (Henderson &
Unwin, 1975; Unwin & Henderson, 1975; Grigorieff et al., 1996; Kühlbrandt et al., 1994; Kimura et al., 1997; Fujiyoshi, 1998; Murata et al., 2000; Gonen et al., 2004, 2005). The first high-resolution model of bacteriorhodopsin to 3.5 Å resolution was quickly
followed by the structure of the light-harvesting chlorophyll complex to 3.4 Å (Kühlbrandt
et al., 1994). However, the application of electron crystallography on thin 3D macromolecular
crystals remained elusive. Early work by Dorset and Parsons (Dorset & Parsons, 1975a, 1975b) demonstrated that electron diffraction data from thin catalase crystals could be
treated kinematically. However, full data sets from each crystal were not collected.
Some of the earliest work on collecting tilt series of thin 3D crystals were performed
on biological crystals (Shi et al., 1998; Jeng & Chiu, 1984; Brink & Chiu, 1994). Seminal work in this area was reported by Wah Chiu and Stokes (Jeng & Chiu, 1984; Shi et al., 1998), predating much of the work on materials, and outlined procedures for collecting
multiple diffraction patterns from a single crystal. While this demonstrated that
this approach was possible, these studies were not capable of determining structures.
It was not until 2013 that of biological material from thin 3D crystals using electron diffraction became a
reality, with the advent of the cryoEM method known as Microcrystal Electron Diffraction
or MicroED (Shi et al., 2013; Nannenga et al., 2014). In MicroED, a crystal is continuously rotated in the electron microscope while
electron diffraction data are collected using a fast camera as a movie. Because electrons
interact with matter 1000× better than X-rays do, the samples for MicroED are much
smaller than for X-ray. The ideal thickness of a crystal for MicroED is about 10–400 nm
for the most commonly used microscopes, thus making a complete diffraction data collection
possible from a crystal a billionth the size of a crystal usually encountered in X-ray
crystallography. This approach solves several problems at once: (1) the effects of
dynamical scattering are efficiently averaged out and reduced from the collection
of the entire reflection intensity profiles during rotation enabling structure solution;
(2) there is no need to orient the crystal prior to data collection, so very low exposures
could be used and biological material could be interrogated; (3) as the experimental
setup is similar to X-ray crystallography, MicroED data could be processed using the
highly sophisticated data reduction software packages that were developed for X-ray
crystallography. This approach meant a major step forward for electron crystallography
and opened the door to many new applications and opportunities, and a wide adoption
of MicroED by the scientific community (Jones et al., 2018; Unge et al., 2024).
Since the first structure using MicroED in 2013, 138 entries of proteins and have been deposited to the Protein Data Bank (PDB) (Berman et al., 2000) as of February 2024, and these include many targets that could only be determined
using MicroED. These targets include, but are not limited to, drugs, natural products, macrocyclic drugs, prions, engineered enzyme variants, ion channels, viral proteins, peptidic antibiotics, amyloid enzymes, and G-Protein Coupled Receptors (GPCRs), demonstrating the general applicability
of MicroED in current structural biology. For many of these samples, the limitations
in crystal growth could not be overcome (Porter et al., 2022; Gillman et al., 2023; Clabbers et al., 2021; Xu et al., 2019) and hindered for years. However, they are ideal for structural investigation using MicroED (Lin
et al., 2023a, 2023b, 2024b; Karothu et al., 2023).
The application of MicroED has also attracted a growing interest for small molecules,
complementing existing techniques with new opportunities. The possibility of using
crystals down to 10 nm in thickness enables samples to be analyzed directly from dry
powder (Bruhn et al., 2021; Sekharan et al., 2021). This circumvents the need to find conditions to grow crystals suitable for single-crystal
X-ray diffraction (SCXRD), which may be difficult for reasons of flexibility, non-specific
contacts, or chemical stability of the compounds developed in modern organic chemistry.
For some of those molecules, recrystallization turned out to be unfruitful or hindered
by sample reactivity, stability, or availability. For several samples, the crystallinity
is low and the ratio of grains with at least decent diffraction may be too low for
a feasible workflow. The resolution to which the crystals diffract may vary between
the grains less than a micrometer in size, which is still doable in the electron beam
of similar width. MicroED may also complement Powder X-ray Diffraction (PXRD), where
can be difficult for mixtures or larger unit cells, where the diffraction peaks overlap.
Using MicroED, single-crystal lattices and phases can be separated and studied individually
(Lightowler et al., 2022; Yokoo et al., 2024), and may include a quantitative analysis of the crystalline content of the powder
(Unge et al., 2024). MicroED is also the method of choice when the amount of sample is limited, for
example, for natural products (Kim et al., 2021a). Its capacity for small crystals has led to the reanalysis and update of the structures
of several pharmacological compounds (Lin et al., 2024a; Kim et al., 2021a, 2021b). The Cambridge Structural Database (CSD; Bruno et al., 2002; Groom et al., 2016), directed towards small molecules, contains more than 740 structures determined
by electron diffraction. Of these, 270 structures state MicroED or the synonymous
keyword cRED (continuous Rotation Electron Diffraction) as the data collection method.
The remaining electron diffraction entries in the CSD are either a result of other
ways to collect diffraction data in an electron microscope or not tagged in a similar
way. Other methods include Precession Electron Diffraction, where the crystallographic
axes tend to be aligned with the beam prior to recording diffraction data, using small
movements of the beam (`precession'). This is more suitable for inorganic material
that is less sensitive to radiation damage due to the extra alignment step. Also included
are data acquisition strategies similar to tomography and X-ray (XFEL) measurements using Serial Electron Diffraction (SerialED). These measurements
sample multiple crystals at a fixed tilt angle to obtain a high data completeness
and distribute the potential radiation damage across the specimen. Due to the conceptually
easy data collection strategy and the similarity to measurements at most synchrotrons,
MicroED is no longer a technique restricted to a few groups. As such it is expected
to continue to grow as a technique, reaching a wider application field. Perhaps the
final hurdle is availability of dedicated MicroED equipment and facilities, which
are still sorely lacking for the scientific community. Once those come online we anticipate
the exponential growth of this field of research to increase even faster.
2. A brief overview of the MicroED workflow
2.1. Sample preparation
MicroED sample preparation generally follows one of two routes depending on whether
the sample is suspended in a liquid or is a dry powder containing submicron crystals.
Solution-immersed samples, such as for macromolecular crystals or small molecule
crystals in suspension, need to be applied to an grid with a thin layer of carbon support. For protein and nucleic acid samples, it
is important that the crystals do not dehydrate, which will severely affect the crystallinity
and therefore the diffraction quality. For these macromolecular samples in aqueous
buffers, excess solution must be blotted away before vitrification in liquid ethane.
For crystalline suspensions of small molecules in solvents with low volatility, the
sample may also need to be blotted to remove excess solvent. If the solvent is highly
volatile, the sample can be allowed to dry before loading into the microscope. Dry
powder samples, on the other hand, can be applied to the grid surface directly. Grains
in a powder that are too large can usually be ground to an appropriate size simply
by grinding the sample between glass microscope slides or using a mortar and pestle.
For crystals that are too thick for beam penetration, cryo-Focused Ion Beam (cryo-FIB)
milling can be used to process the crystals to the proper thickness for MicroED data
collection (Martynowycz et al., 2019). Large crystals can sometimes be disintegrated to small enough crystallites using
techniques frequently exercised for providing seeding fragments in a crystallization
setup or simply by manually cutting the crystal to an appropriate size using dedicated
tools for this purpose (Danelius et al., 2022; Jones et al., 2018; de la Cruz et al., 2017). Each of these samples and methods may have a different grid type which may be ideal,
thus it is important to test several different grid types.
2.2. Data collection
Accurate and careful alignment of the electron beam is necessary for a successful
experiment. The electron beam is easily and accurately bent and controlled using the
magnetic lenses of the electron microscope. As a consequence, any aberrations or imperfections
in the lenses will be directly translated to the resulting diffraction pattern on
the detector, which if uncorrected can lead to distortions of the diffraction pattern.
The built-in deflectors assigned to compensate for these unintentional effects need
to be properly set to ensure an evenly distributed and circular beam on the target.
As the optimal portion of an electron lens is very small, the electron beam also needs
to be aligned accurately onto the optical axis. With a well-aligned electron beam
and a diffraction pattern of decent intensity, the processing of MicroED data is straightforward
using software that was originally intended for X-ray crystallographic diffraction.
Small misalignments of the crystal, beam, detector, or rotation axis are always present,
and the data processing software usually allows for a certain degree of tolerance
to imperfections of the regularity of the diffraction pattern. However, larger deviations
or distortions of the diffraction patterns can lead to errors in the estimation of
the spot position or calculation of the beam profile. This can lead to incorrect integration
of the diffracted intensities (Brázda et al., 2022), resulting in suboptimal data or even failure to process the data.
For successful data collection, the dose (exposure) needs to be calibrated and properly
selected. For protein data collection, it is recommended to use less than 1 e− Å−2 total dose during data collection. The camera speed needs to be optimized in relation
to the rotation speed of the stage. Generally, one should aim to collect the full
rotation range of 140° from each sample to maximize the attainable data. Data collection
then becomes a race against radiation damage. Several reviews and protocols have been
published that delve into the best practices for data collection and integration (Otwinowski
& Minor, 1997; Rajashankar & Dauter, 2014; Gonen, 2013; Hattne et al., 2015).
Because of preferred crystal orientation, low symmetry, and the limitations of the
microscope rotation stage, sometimes the data will be less than 100% complete. These
missing reflections will often lead to an empty and unsampled volume in and an uneven distribution of the measured reflections known as the `missing cone'
(Dorset, 1999). When the missing volume of data is severe, this impedes significantly. To minimize or eliminate the missing cone one could use the suspended
drop approach, as was described recently (Gillman et al., 2024). When crystals do not adopt a collecting data from several crystals allows data sets to be merged and for the to be completely sampled. For example, for both catalase and the SARS-CoV-2 Mpro,
the coverage was drastically enhanced by measuring multiple crystals to reach about
95% completeness. This suggests that even with low crystallographic symmetry, high
completeness can be achieved if data from a number of crystals are merged. If the
crystals do not possess strong preferential orientation, a larger fraction of the
available reflections could be recorded for a better completeness of the merged data.
2.3. Background scattering and noise
The background scattering in MicroED data may be an important consideration when setting
up data collection. The background of the diffraction images is caused by inelastically
scattered electrons from the sample, as well as scattering from the amorphous portion
of the sample, which together results in diffuse scattering. In addition to the background
from the electron beam, the detector will contribute to the noise of the background,
which can be reduced by using newer detectors. Smaller rotation angles per frame can
decrease the amount of background per image when coupled with faster detectors for
These detectors allow for a frame readout speed of hundreds or even thousands of
frames per second and the ability to scan the rotation in greater detail allows better
separation of the diffraction reflections from the background. As a result, the diffraction
data is greatly improved in quality and resolution (Hattne et al., 2023). Additionally, energy filters can remove inelastically scattered electrons and,
if the microscope is equipped with the rather expensive device, can greatly reduce
the amount of background in the data (Clabbers et al., 2025). Using this approach, subatomic resolution was obtained even for proteins.
3. A brief overview of data processing
3.1. Integration
For processing of MicroED data from macromolecules or large unit cells, the same software
that has been developed for X-ray crystallography can be generally used, including
XDS, MOSFLM, DIALS and HKL2000 (Kabsch, 2010a, 2010b; Battye et al., 2011; Parkhurst et al., 2016; Sheldrick, 2015; Otwinowski & Minor, 1997; Clabbers et al., 2018). Small molecule data can be treated in the same way using already familiar software,
including the software used for macromolecules, or using software specifically developed
for electron diffraction, such as PETS2 (Palatinus et al., 2019). While these programs work very well for MicroED data, there are also a few exceptions
to be aware of as a result of the differences between electron and X-ray diffraction
experiments. The wavelength of the electrons typically used in (energies greater than 80 kV) are much shorter than what is used for X-ray diffraction
and the detector distances are significantly longer. Additionally, the lenses of the
electron microscope create aberrations, such as astigmatism, which is more tolerable
in a well-aligned microscope, but the diffraction patterns may still include subtle
distortions that could render the data processing more difficult.
3.2. Scaling and merging
Frequently data from several crystals needs to be merged to increase completeness.
The number of data sets required will depend on the orientation of the crystals and
the crystal symmetry. The standard software for scaling X-ray data is also applicable
for MicroED data. Both POINTLESS/AIMLESS (Evans & Murshudov, 2013) and XSCALE (Kabsch, 2010a) have been used successfully by our research groups, while DIALS also supports the use of several crystals. When merging data from multiple crystals,
it is important to ensure that the correlation between each crystal in the final merged
data set is reasonable; adding several data sets to increase completeness, with a
low correlation to the rest of the data, may ultimately result in worse overall data
for structure solution and refinement.
3.3. Phasing
For macromolecular MicroED structures, the Molecular Replacement (MR) method is usually
a preferred way to phase the data and is often a rapid and easy approach if there
is a good model at hand. The first demonstration of MR in electron diffraction was
reported as early as 2004 (Gonen et al., 2004). If the structure has not been determined previously and there are not any sufficiently
homologous proteins, AlphaFold2 (Jumper et al., 2021) can be used to generate search models. Recently, Alphafold2 was used successfully to phase the MicroED data of a previously unknown protoglobin
target, a variant engineered through directed evolution (Danelius et al., 2022). It has also been applied to phase a protein with unknown function (Miller et al., 2024). The advent of AlphaFold2 promises to expand the applications of MR for MicroED data. In the case of high-resolution
and high-quality data, fragment-based ab initio phasing has also been applied to both peptide (Richards et al., 2021) and protein samples (Martynowycz et al., 2022). For the highest resolution structure, only a generic three amino acid polyalanine
model was required to initiate the phasing of the entire structure. In MR, the chances
of a successful positioning of the model within the is intrinsically dependent on the resolution and completeness, but also on other
quality indicators of the data, such as the signal-to-noise ratio [I/σ(I)]. With MicroED, as with X-ray data, there is not always a clear-cut indication for
when a structure is correctly solved by the MR software, or what quality of data is
needed for a successful In general, a high completeness and high resolution is an advantage, and spending
another few days on the data collection step to improve the data will generally save
significant time for difficult cases in the end.
The structures of smaller molecules, such as pharmaceutical, organic, metal–organic,
or natural compounds, are often approached using ab initio or which provides a structure solution without any prior assumptions. Small molecule
samples regularly result in atomic resolution data and, at the same time, the number
of atoms is smaller and therefore well suited to More recent advances in small molecule structural solution techniques (so-called
Intrinsic Phasing; Sheldrick, 2015) are also applicable. require high-resolution data to resolve atoms (Sheldrick, 1990). The empirically found rule, called the Sheldrick 1.2 Å rule, emerged from the discovery
that a structure was unlikely to be solved unless half the number of theoretically
measurable reflections in the range 1.1–1.2 Å are recorded and have a signal of F > 4σ(F). Granted its importance as a practical confirmation of your data, most entries in
the CSD instead use a lower threshold of about F > 2σ(F) (although variations are plenty). While lacking a correspondence between benchmarks,
including as much data as possible can support the phasing of the structures better
and since the Sheldrick rule was first described, the view of using a signal-to-noise
measure for a resolution cutoff has also shifted. As an example, macromolecular now commonly uses either no sigma cutoff, a lower sigma cutoff, or a CC1/2 cutoff, which makes it more difficult to compare with older standards in general
(Karplus & Diederichs, 2015). For other data-quality indicators, such as completeness and the signal-to-noise
ratio of the data, the threshold for a successful is perhaps not similarly as decisive as the resolution, and the parameters are also
mutually interdependent. For instance, for a phasing approach using with only weaker diffraction data (low signal-to-noise ratio) available, a higher
overall resolution and a completeness close to 100% could be advantageous. Even without
exact thresholds for data-quality indicators, such as completeness or the signal-to-noise
ratio, these properties are nevertheless crucial for a successful ab initio For phasing, the completeness is directly linked to the structural information that
can be extracted from the data. A completeness close to or exceeding 80% would be
the goal as it is more likely a clear structure solution can be found. In contrast,
for data where too low completeness results in a large missing cone of data in one
direction, the systematically missing reflections could lead to the failure of Even if the solution is found with systematic low completeness, the atomic positions
are effectively less resolved in this direction. For phasing, therefore, completeness
is of foremost interest, and a proper scrutinizing process should preceed publishing
structures subject to very low completeness. In our workflow, SHELXT (intrinsic phasing) is often the software of choice for structure solution, which
almost immediately provides the correct structure when data quality is high. As the
chance of greatly increased with high completeness and with data exceeding 3–4 σ to higher resolution, the highest quality data possible should be the goal. There
are other programs that can be used for structure solution, such as SIR2014 (Burla et al., 2015), SUPERFLIP (Palatinus & Chapuis, 2007), and SHELXD (Sheldrick, 2010). Although SHELXD was originally intended for macromolecular phasing through heavy atom or anomalous
peak search, it has been used successfully in the ab initio phasing of small molecule structures. Examples include, for instance, the macrocyclic
compounds romidepsin and paritaprevir, for which data were collected to 0.8 and 0.85 Å,
respectively (Danelius et al., 2023), and two additional crystal forms of the Hepatitis C virus active compound paritaprevir
that were solved to 0.85 and 0.95 Å, respectively (Bu et al., 2024).
In addition to the use of MR on small molecule data has been reported and seems to be a promising
alternative. This is particularly important for samples diffracting to a lower resolution
required by As many samples where MicroED is being used have already been approached using X-rays,
it is often the case that MicroED is pursued as the last resort for difficult samples
and diffraction is known to be compromised. The same software being used for macromolecules
can in principle be used also for small molecules, although in some cases the unit-cell
content calculations will be off, limiting a direct application. To handle the relative
rotation of flexible parts of the molecules, schemes to dissect the molecule in
rigid parts that can be simultaneously searched for have been applied successfully
(Gorelik et al., 2023).
3.4. and model building
of small molecules and large molecules uses the same software as for X-ray crystallography
and with similar settings. SHELXL and Refmac5 have been used successfully to refine small molecule models. Refmac5 (Murshudov et al., 1997) and Phenix (Liebschner et al., 2019) are used frequently for the of macromolecular targets. Just like for X-ray crystallography, there is the risk
of overfitting the model to the data, particularly when data only extend to low resolution
or if the data have low completeness. Similar precautions apply and the Rfree value can help in determining if the model is accurate or overfitted. A difference
between the refinement quality indicators Rwork and Rfree that is substantially larger than 5% is usually an indication of a pushed towards a single indicator (target), and structures reporting these values should be looked at critically.
As MicroED data seem to have a different noise profile, presumably from the differences
in the background and dynamical scattering as compared to X-ray, a higher level of
precaution in terms of overfitting seems sensible. Small molecule data typically
do not have a parameter corresponding to Rfree and overfitting is guarded from by considering the ratio of the number of experimental
observations (reflections) to the number of variable parameters used in the refinement.
Extending the model with additional atoms may be difficult when the phase errors or
lack of high-resolution data inhibits a more detailed map. In that case, it is recommended
to support the placement of atoms during with additional verification of an improved model. As a general rule and with very
few exceptions, the Rfree or R1 value tends to decrease with a better model. For macromolecular data, an omit map
that is calculated without the region of interest is usually calculated to allow
the calculation of a bias-free map. Before the map is calculated with the region of
interest left out, the structure is usually subjected to simulated annealing to reset
the coordinates from any shifts that resulted from bias towards the model. For small
molecule the missing cone can be problematic, at least for the initial rounds of as the exact positions of the atoms are less defined in the direction of the missing
data. In our experience, it may be important to fix the bond lengths and angles to
ideal values early in the process to get better convergence of the parameters. Additionally, the use of anisotropic may not be appropriate with low completeness. As a rule, we generally avoid anisotropic
when the completeness is less than approximately 85% (see Fig. 1 for example).
![[Figure 1]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig1thm.gif) |
Figure 1 The potential maps from MicroED structures with different levels of completeness of data are presented. (a) The refined structure of tyrosine from a commercially purchased powder sample, using overall 96.5% complete data at 0.75 Å resolution. (b) The structure of the long-time used antihistamine Meclizine using 80.7% of possible reflections at 0.96 Å. (c) The structure of valine using 49.4% complete data at 0.75 Å resolution. The missing data, predominantly in one direction (the `missing cone'), results in elongated map densities and less determined coordinates during (d) The output from SHELXT for the same data as in part (c), showing that the missing data causes SHELXT to accidentally include an extra atom in the first model despite the high nominal resolution of the data. |
4. Data and statistics
For evaluation, the structures from both the PDB and the CSD were organized into groups where MicroED was the main method of and groups excluding MicroED, with the vast majority being X-ray structures. The MicroED structures from the PDB that were included in our analysis span a resolution range from 0.75 to 3.4 Å. The structures extracted were included in the statistical analysis only if a minimum number of parameters being extracted was present in the file, including the resolution and Rfree. An initial search for MicroED structures was done using `electron crystallography' for the method label, which also includes structures from 2D crystallography and SerialED. The subset was further refined using both keywords and manual examination of both the PDB header and the primary publication, resulting in the PDB (MicroED) group of structures consisting of 116 structures. This group of MicroED structures was compared with all X-ray-generated structures in the PDB, which were extracted using the same script and the keyword `X-ray', and which generated a much larger group of more than 177000 structures, here referred to as the PDB (X-ray) set. Similarly, MicroED entries were extracted from the CSD using keywords in the comment field expressing either MicroED or cRED, taking variants in upper and lower case into account. Entries where SerialED, Precession Electron Diffraction (PED), or methods using tilt series were mentioned were not used for consistency. In addition, entries were only included if a number of statistical parameters were present to support the quality of the entry, namely, highest/lowest hkl values, goodness-of-fit (GooF), R1(all), R1(I > 2σ), number of reflections above 2σ, number of parameters, and finally the number of restraints in the This left us with 113 structures for analysis, which we denote as the CSD (MicroED) set. For the comparison group, a subset was downloaded manually from the CSD with the same requirements as for the MicroED structures, while excluding all entries mentioning electron diffraction, and are referred to as the CSD (Non-ED) set, consisting of 1023 entries. Entries were selected randomly based on their database ID over the entire name space, except for entries flagged as electron diffraction. For all groups, the statistics were extracted using Python scripts reading the and mmCIF files, as the Python interface from the CCDC could not be used due to lack of access to parameters of interest. Below we discuss the acceptable range and gold standards for each.
4.1. Resolution
In Fig. 2(a), the number of macromolecular MicroED depositions is plotted as distributed over
the entire resolution range. Although this represents a limited set of structures
compared to the similar plot of the X-ray structures in the PDB [Fig. 2(b)], the distribution presents a peak for the number of deposited structures at around
2 Å. At the same time, the distribution for the MicroED structures is more widespread.
This may be a result of the nature of the MicroED analyzed structures (see Conclusion) or variations in the workflow. Interestingly, the peak for the most frequent resolution
range for MicroED structures is also found at 2 Å. This confirms the expectation that
the increased interaction between the electron beam and the crystals would counterweight
the smaller crystal volume. In other words, despite the order of magnitude smaller
crystals utilized in MicroED, the final resolution can be expected to be similar for
MicroED and X-ray data. The resolution is typically not stated explicitly in the files from the CSD and was therefore calculated from the single reflection accepted
of highest resolution using the common cutoff at I > 2σ(I) for the purpose of comparison. The mean resolutions for all structures analyzed
are again very similar between the groups; the CSD (Non-ED) group displays a resolution
of 0.60 Å, compared to 0.64 Å for the CSD (MicroED) group [Table 1, and Figs. 2(c) and 2(d)]. The distribution of the resolution in the CSD is similarly more widespread for
the MicroED group compared to the CSD (Non-ED) group, which has a sharp fall-off at
a resolution of about 1 Å. As a side note, the fact that MicroED contains several
structures at a resolution lower than 1 Å can be considered as an increased demand
for phase-determining techniques, in particular, for MicroED, in addition to the standard
direct methods.
|
![[Figure 2]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig2thm.gif) |
Figure 2 The number of structures plotted for each resolution bin for MicroED structures (blue) and X-ray structures (green). (a) The MicroED structures in the PDB show a widespread distribution with a maximum better than 2 Å as compared to the sharper profile for the X-ray data which is centered at 2 Å (b). The corresponding statistics for the MicroED structures in the CSD (c) are again more widespread compared to the organic Non-ED small molecule structures (d). |
4.2. Completeness and observations over parameters
A completeness of less than 100% can be a consequence of several reasons, such as
limitations in the tilting range of the electron microscope stage, which in the case
of MicroED is an intrinsic property of modern electron microscopes. Larger tilt angles
may also result in higher absorption that could affect the I/σ(I) and the resolution at these angles. In contrast to expectations, as the distribution
of the completeness for all the MicroED depositions in the PDB (MicroED) set plotted
in Fig. 3(a) shows, most structures come from data with a completeness of 70% or higher. It
is common practice to merge data from several MicroED data collections. There are
only a couple of macromolecular MicroED examples that exist where a significantly
lower completeness was reported. While a lower completeness may result in difficulties
in achieving a clear structure solution for both and MR approaches, a high completeness is key to a streamlined process and a reliable structure, as reflected by sound statistical metrics. The
missing cone will have an adverse effect on the final potential map, resulting in
uncertainties of the refined coordinates, as well as possessing a lower effective
resolution in some direction from the systematic lack of data. Considering these
effects, a minimum completeness of 75% should become standard for a given resolution
and structures determined with lower completeness should not be acceptable.
![[Figure 3]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig3thm.gif) |
Figure 3 Statistics of the data collection parameters describing the data consistency of MicroED data (blue) and X-ray data (green). The number of PDB structures for bins of completeness are more distributed for the MicroED data (a) compared to the X-ray data (b). This can be interpreted as a result of the limitations of the available tilting range in an electron microscope setup. In (c), the distribution of CC1/2 is plotted, revealing a typical value of above 0.85 with variations for the MicroED data and with more consistent values in the case of the X-ray data (d). The distribution of Rmerge is plotted for MicroED data (e) and X-ray data (f) and is, together with CC1/2, related to the internal consistency of the collected intensities. Larger differences can be seen in the MicroED data, however, the final refinements of the structures display similar results in terms of data quality. |
The eligibility of the of small molecules is supported by the number of observed reflections and the number
of parameters, as well as the number of restraints used in the While originally a ratio of 4 for the number of reflections to the number of parameters
was considered enough, currently a ratio of 10 or more is deemed appropriate to support
the of the model as described in the checkCIF routine REFNR01. For comparison, the `refinement redundancy' is calculated and presented in Table 2, and defined as:
|
||||||||||||||||||||||||||||||||
#ref = Redundancy in = (No. of reflections + No. of restrictions)/No. of parameters [I > 2σ(I)]
As there is no consensus of an appropriate ratio with respect to MicroED data, we suggest the inclusion of the ratio, as well as the actual number of observations/restrictions and parameters refined, with the published structure. For the subset used from the CSD, the redundancy ratios are 15.7 for the MicroED set and 16.4 for the electron diffraction excluded set. With both around 16, they are well above the reliable bar of 10. Not surprisingly, considering the similarity of the methods, we can confirm similar values for MicroED and X-ray crystallography.
4.3. Merging R factors
To compensate for the limitations in the stage rotation range and weaker diffraction
at higher tilt angles, data from several crystals will frequently be merged to increase
the completeness for and the calculation of improved maps. Merging several data sets is common and works
well for isomorphous data sets, but merging may be a non-trivial task for data sets
not sufficiently isomorphous. Therefore, it has often become routine to identify groups
of data sets that result in coherent structure factors, as determined by the correlation
coefficients or any R factor related to merging, such as Rmerge, Rpim, or Rmeas. The Rmerge value can, for MicroED data, sometimes be substantially higher than expected, between
symmetry-related reflections within the data set, and between data from similar crystals
there may be large variations. In the analyzed macromolecular groups displayed in
Fig. 3(c), almost 70% of the data displays an Rmerge value of 30% or lower, and most of the reported structures are found in the peak
at 15–30% in Rmerge. With these numbers, which are also not unusual for X-ray data, especially for the
high-resolution shells, the influence of non-isomorphism between crystals seems not
to be a major concern. A smaller portion of the data resulted in an unusually high
Rmerge value between 50 and 77% (Lanza et al., 2019; Takaba et al., 2021; Martynowycz et al., 2021; de la Cruz et al., 2017). Nevertheless, even the data with rather high Rmerge also resulted in well-refined structures and interpretable maps. In summary, based
on the reported work, Rmerge can be expected to fall in the region of 35% or lower, although it seems that a reasonable
structure can occasionally be achieved for data that does not reach the expected levels.
For small molecule data, merging R values between data sets are often not reported.
4.4. CC1/2
Karplus and Diederichs suggested in 2012 that a Pearson between two subsets of the total data, CC1/2, is a better indicator of the quality and resolution of crystallographic data sets
than more traditional measures (Karplus & Diederichs, 2012, 2015). Specifically, it was suggested that the CC1/2 could be analyzed in resolution bins to verify the extent of the diffracting pattern.
It has been argued, as well as becoming a standard for many, that data in a resolution
shell with a CC1/2 below 0.5 could still be used for structure towards X-ray crystallographic data. For the subset of our analysis, the published
MicroED structures report a high overall CC1/2 – better than 80% for all structures and better than 90% only with a few exceptions
[Fig. 3(e)]. It is suggested that a CC1/2 of at least 90% for the overall data could be considered standard, while there may
be less congruence in the value in the highest resolution bin.
4.5. Residual R factors
It has been commonly anticipated that many of the structures determined by electron
diffraction display residual R factors (R1, Rwork, and Rfree for small molecules and macromolecule respectively) that tend to be higher for MicroED data in comparison with X-ray data.
As the distribution of the refined Rfree for all X-ray structures in the PDB shows (Fig. 4), most deposited structures refine to an Rfree value in the range 18–28%. The corresponding plot for MicroED indicates that a similar
peak can be found at about 20–30% in Rfree, only about 2% higher in general. Similarly, the distribution of the Rwork value displays a peak at 16–22% for the X-ray structures and a corresponding peak
only slightly higher than for MicroED, at 18–24% in Rwork. While MicroED results in slightly higher R values, certain standards should be appropriate. Almost all MicroED structures, regardless
of resolution, display Rwork values less than 30% and for Rfree less than 35%. As these numbers have been consistent in our laboratories, we believe
they could serve as a standard goal for other MicroED studies. Plotted as a trend
of the resolution (Figs. 5 and 6), the R values are generally improved for structures refined to a higher resolution, and
structures with a resolution of 1.5 Å or better are generally found to achieve an
Rwork value of 25%. The trends of a lower Rfree value at higher resolution can be understood as a better agreement of the model as
the data becomes more well determined. Overall, the gap between X-ray and MicroED
structures in terms of Rwork and Rfree is generally a few percent. In the small molecule files, the R1(all) is the conventional R factor within the given resolution interval for all data. The MicroED mean value
of 0.23 for R1(all) is greater than similar values for non-MicroED data of 0.08. As small molecule
structural models typically include all available atoms, the discrepancy points to
differences in the structure factors or calculations. As with the resolution (see Conclusions), both Rfree and R1(all) have a wider distribution, which may be a result of sample selection, as discussed
below. It may also be contributed to by differences in the structure factors, such
as residuals of dynamical scattering and a comparably higher degree of inelastic background
scattering in the MicroED data. In summary, there is no reason to believe that the
difference in the residuals is related to not being able to model part of the structure;
the models resulting from MicroED seem to contain a similar degree of detail as with
X-ray-generated data. We anticipate that the slightly higher R values result from properties in the data, which as a result may contribute to some
higher level of noise in the potential maps, but it is not related to properties of
the final structure. An accurately refined model should not be of any concern for
most applications.
![[Figure 4]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig4thm.gif) |
Figure 4 Statistics of the structure refinement parameters for macromolecular MicroED structures (blue) and X-ray structures (green) from the PDB (a)–(d) and the CSD (e)/(f). The distribution of both Rfree (a) and Rwork (b) are similar for the macromolecular MicroED and X-ray structures, with a slight shift towards higher values of about 2% for the MicroED data. Although interesting from the point of consistency within the PDB, the negligible difference also confirms the similar quality of the resulting structures for the two methods. The corresponding distribution of R1(all) in the CSD displays a larger difference of ∼0.15%. Small molecule models are generally more complete than macromolecules, generating smaller R values; hence discrepancies between electron and non-electron data stand out better. Additionally, the larger number of reflections typically measured in macromolecular data might to some extent smear out the effect of dynamical scattering, which is affected by strong reflections and also varies between close reflections taking a similar trajectory through the crystal. |
![[Figure 5]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig5thm.gif) |
Figure 5 Rfree plotted as a function of resolution, showing the macromolecular MicroED structures (blue points) and X-ray structures (green points). The trend lines are shown for the distribution of the PDB MicroED and X-ray data (blue and green lines, respectively). Although the trend lines for the MicroED and X-ray structures follow each other closely, Rfree tends to be somewhat greater, in particular for the higher resolution structures. For visibility, 2% of the X-ray data was randomly selected for presentation. |
![[Figure 6]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig6thm.gif) |
Figure 6 The R1(all) plotted as a function of resolution, showing the MicroED structures in the CSD group (MicroED set; blue points) and the comparative group (Non-ED group; green points). The trend lines are shown for the distribution of the MicroED and X-ray data (blue and green lines respectively). |
4.6. Goodness-of-fit (GooF)
The goodness-of-fit parameter takes the difference in structure factors and the number of observed reflections into account, and also the number of parameters used. At the end of a the GooF is expected to approach 1 and is mostly used for small molecules. The fact that the mean value of 1.06 is closer to the ideal value of 1.0 for the CSD (Non-ED) group than the mean value of 1.7 for the CSD (MicroED) data may be for reasons similar to the differences in the structure factors, as discussed in previous sections. As with the resolution, the values for MicroED data contribute to a larger variety, with several structures reaching higher values.
4.7. Atomic displacement parameters
The term Atomic Displacement Parameter or Anisotropic Displacement Parameter (ADP)
has unfortunately been used in an inconsistent way historically. In the macromolecular
world, it is often meant to refer to the B factor, while in the small molecule literature it is referring to the Uij. Both have the same dimension of Å2 describing the same phenomenon and are related by B = 8π2U. The B factor was originally referred to as the temperature factor or the Debye–Waller factor
and was used to model the attenuation of diffraction intensities due to thermal motion,
but later ADPs also became a way to model atom displacement and conformational or
crystal disorder with a similar influence on the diffraction intensities. ADPs therefore
provide a wide range of information, such as atomic motion, which are usually short
distances, protein conformation disorder, as well as protein structure dynamics, which
are typically large-scale movements. B factor values that are too large, for instance more than 100 Å2, are generally not considered to have any physical meaning (Carugo, 2018a, 2018b). Instead, their usage in structure may lead to overfitting or mask an error of the choice of atom type, in particular
for non-covalently bound atoms. Very high ADPs can sometimes be a motivation to instead
use more descriptive attributes, such as a lower occupancy for a group of atoms, where
alternative positions would be a better description.
The overall B is correlated with the resolution and within a non-redundant subset of the PDB it is found that:
B = 9*resolution2 (Å2)
For instance, a model determined to a high resolution of 1 Å has an overall B closer to 10 Å2, but for medium or lower resolution structures, higher B factors are common. In general, the overall B falls between 10 and 30 Å2 for most structures, while an B of lower than 10 Å2 or substantially higher than 50 Å2 is less frequent. Instead, these less likely overall ADPs may indicate a problem with the model (or data), although exceptions do exist. An unreasonable ADP can mask an atom at the wrong position and it is important to confirm for any atom in doubt that its position is reasonable in the structure. During model building and it is useful to confirm a deviating ADP by checking the closest environment, such as surrounding charged or hydrophobic atoms, and type of coordinating atoms.
In an analysis of the ADPs of protein structures, Carugo observed that for X-ray structures,
a resolution of 1.5 Å or better resulted in average B factors of about 25 and only at lower resolution did they tend to increase. As the
resolution improves, the B factor decreases and its extrapolation would intersect with the resolution axis
at about 0.5 Å. The plot of the model B factors versus the resolution of published MicroED structures shows an almost identical behavior:
almost all mean temperature factors are below 20 Å2 at a resolution of 1.5 Å and the extrapolated tendency would intersect the resolution
axis at around 0.5 Å or better (Fig. 7). The only noticeable difference is that the B factors are somewhat lower for the MicroED subset. However, given the limited number
of MicroED data sets in the comparison, it is difficult to speculate whether this
is significant or if there are any reasons for the slightly lower B factor in the MicroED subset. The B factors of the structures in the PDB are far more distributed, some reaching very
high values, whereas the numbers converge more as the resolution get better.
![[Figure 7]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig7thm.gif) |
Figure 7 Overall B factors plotted as a function of resolution, showing the MicroED structures (blue points) and the X-ray structures (green points). The trend lines are shown for the distribution of the MicroED and X-ray data (blue and green lines, respectively). The overall B factor is substantially lower for the MicroED structures, in particular at lower resolution. This may partially reflect the larger variation of structures solved using X-ray, but also demonstrates the high quality of the atomic coordinates in the of the MicroED structures. For visibility, 2% of the X-ray data was randomly selected for presentation. |
4.8. Summary of quality indicators
Despite efforts to prepare comparative groups of structures for MicroED and Non-MicroED structures, there will be differences from factors outside of this scope. Several structures determined by MicroED are targets used as test samples for evaluating the method and therefore are expected to form well-behaving crystals, while many others represent a group of structures that have been unusually difficult to analyze using X-ray crystallography and where crystal growth was likely very limited, calling for the use of MicroED for these samples. It is probable that both the well-behaved and the limited crystal growth samples are reflected in the nanocrystals used in MicroED. A likely outcome could be the drawn-out nature of the statistical analysis. While MicroED also uses software tuned for X-ray crystallography, it is possible that MicroED data analysis workflows are less homogeneous compared with X-ray data, thus leading to a larger statistical range of refined models. As a main observation, it is illustrative to note that the parameters Rmerge and CC1/2 are the only parameters in this analysis where clearer differences between MicroED and X-ray data can be discerned for structures in the PDB. These parameters are related to the internal consistency of the measured intensities and describe the data before merging of the intensities and multiple data sets. The quality of the final structures, after merging of data and of the structures, are better addressed by parameters from the and structure evaluation. The take-away of these frequent comparisons to X-ray structures is that, while the statistical analysis of MicroED data may seem to be a matter of concern at the data collection stage, the quality of the resulting structures after data merging and compares very well between MicroED and X-ray crystallography – as would be expected from substantially related techniques. For instance, the resolution ends up with identical peaks for the most frequent resolution intervals, and the R factors are overall similar and differ by just a couple of percent. While the resolution is strongly associated with the properties of the crystals, the Rmerge and CC1/2 values are directly related to the internal consistency of the data, which may depend on the data collection setup using an electron microscope, as well as the interaction of the electron beam with the crystals. The residuals on the other hand are related to the data only after merging. Agreement between the residuals for MicroED and X-ray data reflects visually comparable outcomes of the two methods for macromolecular and a small but consistent difference in the case of small molecule which may be minimized taking dynamical scattering into account during This observation, however, neglects the differences in the sizes of the crystals. Overall, the resolution of a data collection using MicroED can be comparable to the resolution from data from X-ray crystallography data collection even though using crystals orders of magnitude smaller.
5. Conclusions
More than a decade after the first MicroED structure was deposited and the method unveiled, there are more than a hundred macromolecular and several hundred small molecule structures deposited in the PDB and CSD, respectively. As the number of refined structures with their data become available to the community, there are lessons that can be learned with respect to the profiles and quality of MicroED structures, as well as best practices. Here we have summarized the lessons learned from these data and proposed gold standards to be implemented by the community. Generally, the resulting models and density maps from MicroED are of high quality and comparable in resolution to X-ray crystallography. At this stage, the biggest hurdle for an even wider dissemination of the MicroED technology to the scientific community is the lack of experience, infrastructure, and national facilities with robust dedicated instruments for the method. As the tools for MicroED improve, and more cost-effective automatic microscopes for MicroED are deployed, we expect the use of MicroED to continue its rapid growth.
![[Figure 8]](https://journals.iucr.org/10.1107/S2053229625004875/nh3003fig8thm.gif) |
Figure 8 The distribution of goodness-of-fit (GooF) plotted as a function of resolution, showing the MicroED structures in the CSD group (MicroED set; blue points) and the comparative group (Non-ED group; green points). The lines with similar colors mark the mean values of the entire interval. Outliers are not included for visibility. |
Acknowledgements
We acknowledge the kind support from Stockholm University. This study was supported by Stiftelsen för Strategisk Forskning – Research Infrastructure 2 (SSF-RIF2), the National Institutes of Health, and the Department of Defense. The Gonen Laboratory is supported by funds from the Howard Hughes Medical Institute. Portions of this research or manuscript completion were developed with funding from the Department of Defense. Effort sponsored by the US Government (under Other Transaction number W15QKN-16-9-1002) between the MCDC, and the Government. The US Government is authorized to reproduce and distribute reprints for Governmental purposes, notwithstanding any copyright notation thereon. The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies or endorsements, either expressed or implied, of the US Government. The PAH shall flow down these requirements to its sub-awardees, at all tiers.
Conflict of interest
The authors declare no competing interests.
Funding information
Funding for this research was provided by: National Institute of General Medical Sciences (grant No. P41GM136508 to Tamir Gonen); Department of Defense (grant No. HDTRA1-21-1-0004; grant No. MCDC-2202-002).
References
Andrusenko, I., Krysiak, Y., Mugnaioli, E., Gorelik, T. E., Nihtianova, D. & Kolb,
U. (2015). Acta Cryst. B71, 349–357. CrossRef ICSD IUCr Journals Google Scholar
Battye, T. G. G., Kontogiannis, L., Johnson, O., Powell, H. R. & Leslie, A. G. W.
(2011). Acta Cryst. D67, 271–281. Web of Science CrossRef CAS IUCr Journals Google Scholar
Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H., Shindyalov,
I. N. & Bourne, P. E. (2000). Nucleic Acids Res. 28, 235–242. Web of Science CrossRef PubMed CAS Google Scholar
Brázda, P., Klementová, M., Krysiak, Y. & Palatinus, L. (2022). IUCrJ, 9, 735–755. Web of Science CrossRef PubMed IUCr Journals Google Scholar
Brink, J. & Chiu, W. (1994). J. Struct. Biol. 113, 23–34. CrossRef CAS PubMed Web of Science Google Scholar
Bruhn, J. F., Scapin, G., Cheng, A., Mercado, B. Q., Waterman, D. G., Ganesh, T.,
Dallakyan, S., Read, B. N., Nieusma, T., Lucier, K. W., Mayer, M. L., Chiang, N. J.,
Poweleit, N., McGilvray, P. T., Wilson, T. S., Mashore, M., Hennessy, C., Thomson,
S., Wang, B., Potter, C. S. & Carragher, B. (2021). Front. Mol. Biosci. 8, https://doi.org/10.3389/fmolb.2021.648603. Google Scholar
Bruno, I. J., Cole, J. C., Edgington, P. R., Kessler, M., Macrae, C. F., McCabe, P.,
Pearson, J. & Taylor, R. (2002). Acta Cryst. B58, 389–397. Web of Science CrossRef CAS IUCr Journals Google Scholar
Bu, G., Danelius, E., Wieske, L. H. E. & Gonen, T. (2024). Adv. Biol. 8, 2300570. Google Scholar
Burla, M. C., Caliandro, R., Carrozzini, B., Cascarano, G. L., Cuocci, C., Giacovazzo,
C., Mallamo, M., Mazzone, A. & Polidori, G. (2015). J. Appl. Cryst. 48, 306–309. Web of Science CrossRef CAS IUCr Journals Google Scholar
Carugo, O. (2018a). Amino Acids, 50, 775–786. Web of Science CrossRef CAS PubMed Google Scholar
Carugo, O. (2018b). BMC Bioinformatics, 19, 61. Google Scholar
Clabbers, M. T. B., Gruene, T., Parkhurst, J. M., Abrahams, J. P. & Waterman, D. G.
(2018). Acta Cryst. D74, 506–518. Web of Science CrossRef IUCr Journals Google Scholar
Clabbers, M. T. B., Hattne, J., Martynowycz, M. W. & Gonen, T. (2025). Nat. Commun. 16, 2247. PubMed Google Scholar
Clabbers, M. T. B., Holmes, S., Muusse, T. W., Vajjhala, P. R., Thygesen, S. J., Malde,
A. K., Hunter, D. J. B., Croll, T. I., Flueckiger, L., Nanson, J. D., Rahaman, Md.
H., Aquila, A., Hunter, M. S., Liang, M., Yoon, C. H., Zhao, J., Zatsepin, N. A.,
Abbey, B., Sierecki, E., Gambin, Y., Stacey, K. J., Darmanin, C., Kobe, B., Xu, H.
& Ve, T. (2021). Nat. Commun. 12, 2578. Web of Science CrossRef PubMed Google Scholar
Cowley, J. M. (1953). Acta Cryst. 6, 516–521. CrossRef CAS IUCr Journals Web of Science Google Scholar
Cowley, J. M. (1956). Acta Cryst. 9, 391–396. CrossRef ICSD CAS IUCr Journals Google Scholar
Danelius, E., Bu, G., Wieske, H. & Gonen, T. (2023). BioRxiv, https://doi.org/10.1101/2023.07.31.551405. Google Scholar
Danelius, E., Porter, N. J., Unge, J., Arnold, F. H. & Gonen, T. (2022). BioRxiv, https://doi.org/10.1101/2022.10.18.512604. Google Scholar
de la Cruz, M. J., Hattne, J., Shi, D., Seidler, P., Rodriguez, J., Reyes, F. E.,
Sawaya, M. R., Cascio, D., Weiss, S. C., Kim, S. K., Hinck, C. S., Hinck, A. P., Calero,
G., Eisenberg, D. & Gonen, T. (2017). Nat. Methods, 14, 399–402. CAS PubMed Google Scholar
Dorset, D. L. (1999). Microsc. Res. Tech. 46, 98–103. PubMed CAS Google Scholar
Dorset, D. L., Jap, B. K., Ho, M.-H. & Glaeser, R. M. (1979). Acta Cryst. A35, 1001–1009. CrossRef CAS IUCr Journals Web of Science Google Scholar
Dorset, D. L. & Parsons, D. F. (1975a). Acta Cryst. A31, 210–215. CrossRef CAS IUCr Journals Web of Science Google Scholar
Dorset, D. L. & Parsons, D. F. (1975b). J. Appl. Cryst. 8, 12–14. CrossRef IUCr Journals Web of Science Google Scholar
Evans, P. R. & Murshudov, G. N. (2013). Acta Cryst. D69, 1204–1214. Web of Science CrossRef CAS IUCr Journals Google Scholar
Fujiyoshi, Y. (1998). Adv. Biophys. 35, 25–80. CrossRef CAS PubMed Google Scholar
Gillman, C., Bu, G., Danelius, E., Hattne, J., Nannenga, B. & Gonen, T. (2024). BioRxiv, https://doi.org/10.1101/2024.01.11.575283. Google Scholar
Gillman, C., Patel, K., Unge, J. & Gonen, T. (2023). BioRxiv, https://doi.org/10.1101/2023.03.31.535166. Google Scholar
Gonen, T. (2013). Electron Crystallography of Soluble and Membrane Proteins: Methods and Protocols, edited by I. Schmidt-Krey & Y. Cheng, pp. 153–169. Totowa, NJ: Humana Press. Google Scholar
Gonen, T., Cheng, Y., Sliz, P., Hiroaki, Y., Fujiyoshi, Y., Harrison, S. C. & Walz,
T. (2005). Nature, 438, 633–638. Web of Science CrossRef PubMed CAS Google Scholar
Gonen, T., Sliz, P., Kistler, J., Cheng, Y. & Walz, T. (2004). Nature, 429, 193–197. Web of Science CrossRef PubMed CAS Google Scholar
Gorelik, T. E., Lukat, P., Kleeberg, C., Blankenfeldt, W. & Mueller, R. (2023). Acta Cryst. A79, 504–514. CSD CrossRef IUCr Journals Google Scholar
Grigorieff, N., Ceska, T. A., Downing, K. H., Baldwin, J. M. & Henderson, R. (1996).
J. Mol. Biol. 259, 393–421. CrossRef CAS PubMed Web of Science Google Scholar
Groom, C. R., Bruno, I. J., Lightfoot, M. P. & Ward, S. C. (2016). Acta Cryst. B72, 171–179. Web of Science CrossRef IUCr Journals Google Scholar
Hattne, J., Clabbers, M. T. B., Martynowycz, M. W. & Gonen, T. (2023). BioRxiv, https://doi.org/10.1101/2023.06.29.547123. Google Scholar
Hattne, J., Reyes, F. E., Nannenga, B. L., Shi, D., de la Cruz, M. J., Leslie, A.
G. W. & Gonen, T. (2015). Acta Cryst. A71, 353–360. Web of Science CrossRef IUCr Journals Google Scholar
Henderson, R. & Unwin, P. N. (1975). Nature, 257, 28–32. PubMed CAS Google Scholar
Jeng, T.-W. & Chiu, W. (1984). Ultramicroscopy, 13, 27–34. CrossRef CAS PubMed Web of Science Google Scholar
Jones, C. G., Martynowycz, M. W., Hattne, J., Fulton, T. J., Stoltz, B. M., Rodriguez,
J. A., Nelson, H. M. & Gonen, T. (2018). ACS Cent. Sci. 4, 1587–1592. Web of Science CSD CrossRef CAS PubMed Google Scholar
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool,
K., Bates, R., Žídek, A., Potapenko, A., Bridgland, A., Meyer, C., Kohl, S. A. A.,
Ballard, A. J., Cowie, A., Romera-Paredes, B., Nikolov, S., Jain, R., Adler, J., Back,
T., Petersen, S., Reiman, D., Clancy, E., Zielinski, M., Steinegger, M., Pacholska,
M., Berghammer, T., Bodenstein, S., Silver, D., Vinyals, O., Senior, A. W., Kavukcuoglu,
K., Kohli, P. & Hassabis, D. (2021). Nature, 596, 583–589. Web of Science CrossRef CAS PubMed Google Scholar
Kabsch, W. (2010a). Acta Cryst. D66, 133–144. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kabsch, W. (2010b). Acta Cryst. D66, 125–132. Web of Science CrossRef CAS IUCr Journals Google Scholar
Karothu, D. P., Alhaddad, Z., Göb, C. R., Schürmann, C. J., Bücker, R. & Naumov, P.
(2023). Angew. Chem. Int. Ed. 62, e202303761. Web of Science CSD CrossRef Google Scholar
Karplus, P. A. & Diederichs, K. (2012). Science, 336, 1030–1033. Web of Science CrossRef CAS PubMed Google Scholar
Karplus, P. A. & Diederichs, K. (2015). Curr. Opin. Struct. Biol. 34, 60–68. Web of Science CrossRef CAS PubMed Google Scholar
Kim, L. J., Ohashi, M., Zhang, Z., Tan, D., Asay, M., Cascio, D., Rodriguez, J. A.,
Tang, Y. & Nelson, H. M. (2021a). Nat. Chem. Biol. 17, 872–877. CAS PubMed Google Scholar
Kim, L. J., Xue, M., Li, X., Xu, Z., Paulson, E., Mercado, B., Nelson, H. M. & Herzon,
S. B. (2021b). J. Am. Chem. Soc. 143, 6578–6585. CAS PubMed Google Scholar
Kimura, Y., Vassylyev, D. G., Miyazawa, A., Kidera, A., Matsushima, M., Mitsuoka,
K., Murata, K., Hirai, T. & Fujiyoshi, Y. (1997). Nature, 389, 206–211. CrossRef CAS PubMed Web of Science Google Scholar
Kühlbrandt, W., Wang, D. N. & Fujiyoshi, Y. (1994). Nature, 367, 614–621. PubMed Web of Science Google Scholar
Lanza, A., Margheritis, E., Mugnaioli, E., Cappello, V., Garau, G. & Gemmi, M. (2019).
IUCrJ, 6, 178–188. Web of Science CrossRef CAS PubMed IUCr Journals Google Scholar
Liebschner, D., Afonine, P. V., Baker, M. L., Bunkóczi, G., Chen, V. B., Croll, T.
I., Hintze, B., Hung, L.-W., Jain, S., McCoy, A. J., Moriarty, N. W., Oeffner, R.
D., Poon, B. K., Prisant, M. G., Read, R. J., Richardson, J. S., Richardson, D. C.,
Sammito, M. D., Sobolev, O. V., Stockwell, D. H., Terwilliger, T. C., Urzhumtsev,
A. G., Videau, L. L., Williams, C. J. & Adams, P. D. (2019). Acta Cryst. D75, 861–877. Web of Science CrossRef IUCr Journals Google Scholar
Lightowler, M., Li, S., Ou, X., Zou, X., Lu, M. & Xu, H. (2022). Angew. Chem. Int. Ed. 61, e202114985. Web of Science CSD CrossRef Google Scholar
Lin, J., Bu, G., Unge, J. & Gonen, T. (2024a). BioRxiv, https://doi.org/10.1101/2024.06.05.597682. Google Scholar
Lin, J., Bu, G., Unge, J. & Gonen, T. (2024b). BioRxiv, https://doi.org/10.1101/2024.01.04.574265. Google Scholar
Lin, J., Unge, J. & Gonen, T. (2023a). BioRxiv, https://doi.org/10.1101/2023.06.28.546957. Google Scholar
Lin, J., Unge, J. & Gonen, T. (2023b). BioRxiv, https:/doi.org/10.1101/2023.09.05.556418. Google Scholar
Martynowycz, M. W., Clabbers, M. T. B., Hattne, J. & Gonen, T. (2022). Nat. Methods, 19, 724–729. Web of Science CrossRef CAS PubMed Google Scholar
Martynowycz, M. W., Clabbers, M. T. B., Unge, J., Hattne, J. & Gonen, T. (2021). Proc. Natl. Acad. Sci. USA, 118, e210884118. Google Scholar
Martynowycz, M. W., Zhao, W., Hattne, J., Jensen, G. J. & Gonen, T. (2019). Structure, 27, 545–548. Web of Science CrossRef CAS PubMed Google Scholar
Miller, J. E., Agdanowski, M. P., Dolinsky, J. L., Sawaya, M. R., Cascio, D., Rodriguez,
J. A. & Yeates, T. O. (2024). Acta Cryst. D80, 270–278. CrossRef IUCr Journals Google Scholar
Murata, K., Mitsuoka, K., Hirai, T., Walz, T., Agre, P., Heymann, J. B., Engel, A.
& Fujiyoshi, Y. (2000). Nature, 407, 599–605. Web of Science CrossRef PubMed CAS Google Scholar
Murshudov, G. N., Vagin, A. A. & Dodson, E. J. (1997). Acta Cryst. D53, 240–255. CrossRef CAS Web of Science IUCr Journals Google Scholar
Nannenga, B. L., Shi, D., Leslie, A. G. W. & Gonen, T. (2014). Nat. Methods, 11, 927–930. Web of Science CrossRef CAS PubMed Google Scholar
Oleynikov, P., Hovmöller, S. & Zou, X. D. (2007). Ultramicroscopy, 107, 523–533. Web of Science CrossRef PubMed CAS Google Scholar
Otwinowski, Z. & Minor, W. (1997). Methods in Enzymology, Vol. 276, Macromolecular Crystallography, Part A, edited by C. W. Carter Jr & R. M. Sweet, pp. 307–326. New York: Academic
Press. Google Scholar
Palatinus, L., Brázda, P., Jelínek, M., Hrdá, J., Steciuk, G. & Klementová, M. (2019).
Acta Cryst. B75, 512–522. Web of Science CrossRef IUCr Journals Google Scholar
Palatinus, L. & Chapuis, G. (2007). J. Appl. Cryst. 40, 786–790. Web of Science CrossRef CAS IUCr Journals Google Scholar
Parkhurst, J. M., Winter, G., Waterman, D. G., Fuentes-Montero, L., Gildea, R. J.,
Murshudov, G. N. & Evans, G. (2016). J. Appl. Cryst. 49, 1912–1921. Web of Science CrossRef CAS IUCr Journals Google Scholar
Porter, N. J., Danelius, E., Gonen, T. & Arnold, F. H. (2022). J. Am. Chem. Soc. 144, 8892–8896. CAS PubMed Google Scholar
Rajashankar, K. & Dauter, Z. (2014). Structural Genomics and Drug Discovery: Methods and Protocols, edited by W. F. Anderson, pp. 211–237. New York: Springer. Google Scholar
Richards, L. S., Flores, M. D., Millán, C., Glynn, C., Zee, C.-T., Sawaya, M. R.,
Gallagher-Jones, M., Borges, R. J., Usón, I. & Rodriguez, J. A. (2021). BioRxiv, https://doi.org/10.1101/2021.09.13.459692. Google Scholar
Sekharan, S., Liu, X., Yang, Z., Liu, X., Deng, L., Ruan, S., Abramov, Y., Sun, G.,
Li, S., Zhou, T., Shi, B., Zeng, Q., Zeng, Q., Chang, C., Jin, Y. & Shi, X. (2021).
RSC Adv. 11, 17408–17412. Web of Science CSD CrossRef CAS PubMed Google Scholar
Sheldrick, G. M. (1990). Acta Cryst. A46, 467–473. CrossRef CAS Web of Science IUCr Journals Google Scholar
Sheldrick, G. M. (2010). Acta Cryst. D66, 479–485. Web of Science CrossRef CAS IUCr Journals Google Scholar
Sheldrick, G. M. (2015). Acta Cryst. A71, 3–8. Web of Science CrossRef IUCr Journals Google Scholar
Shi, D., Lewis, M. R., Young, H. S. & Stokes, D. L. (1998). J. Mol. Biol. 284, 1547–1564. Web of Science CrossRef CAS PubMed Google Scholar
Shi, D., Nannenga, B. L., Iadanza, M. G. & Gonen, T. (2013). Elife, 2, e01345. Web of Science CrossRef PubMed Google Scholar
Takaba, K., Maki-Yonekura, S., Inoue, S., Hasegawa, T. & Yonekura, K. (2021). Front. Mol. Biosci. 7, https://doi.org/10.3389/fmolb.2020.612226. Google Scholar
Unge, J., Lin, J., Weaver, S. J., Sae Her, A. & Gonen, T. (2024). Adv. Sci. 11, 2400081. Google Scholar
Unwin, P. N. T. & Henderson, R. (1975). J. Mol. Biol. 94, 425–440. CrossRef PubMed CAS Web of Science Google Scholar
Vainshtein, B. K. & Pinsker, Z. G. (1949). J. Phys. Chem. SSSR, 23, 1058–1067. Google Scholar
Vincent, R. & Midgley, P. A. (1994). Ultramicroscopy, 53, 271–282. CrossRef CAS Web of Science Google Scholar
Xu, H., Lebrette, H., Clabbers, M. T. B., Zhao, J., Griese, J. J., Zou, X. & Högbom,
M. (2019). Sci. Adv. 5, eaax4621. Web of Science CrossRef PubMed Google Scholar
Yokoo, H., Aoyama, Y., Matsumoto, T., Yamamoto, E., Uchiyama, N. & Demizu, Y. (2024).
Chem. Pharm. Bull. 72, 471–474. CAS Google Scholar
This is an open-access article distributed under the terms of the Creative Commons Attribution (CC-BY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.
|
STRUCTURAL CHEMISTRY |
