research papers
Similarity score for screening phaseretrieved maps in Xray diffraction imaging – characterization in reciprocal space
^{a}Graduate School of Science, University of Hyogo, 321 Kouto, Kamigoricho, Akogun, Hyogo 6781297, Japan, ^{b}RIKEN SPring8 Center, 111 Kouto, Sayo, Sayogun, Hyogo 6795148, Japan, ^{c}Synchrotron Radiation Research Center, Hyogo Science and Technology Association, 14902 Kouto, Shingu, Tatsuno, Hyogo 6795148, Japan, ^{d}International Center for Synchrotron Radiation Innovation Smart, Tohoku University, Katahira 211, Aobaku, Sendai 9808577, Japan, and ^{e}Department of Physics, Faculty of Science and Technology, Keio University, 3141 Hiyoshi, Kohokuku, Yokohama, Kanagawa 2238522, Japan
^{*}Correspondence email: nakasako@phys.keio.ac.jp
Xray diffraction imaging (XDI) is utilized for visualizing the structures of noncrystalline particles in material sciences and biology. In the structural analysis, phaseretrieval (PR) algorithms are applied to the diffraction amplitude data alone to reconstruct the
of a specimen particle projected along the direction of the incident Xrays. However, PR calculations may not lead to good convergence because of a lack of diffraction patterns in smallangle regions and Poisson noise in Xray detection. Therefore, the PR calculation is still a bottleneck for the efficient application of XDI in the structural analyses of noncrystalline particles. For screening maps from hundreds of trial PR calculations, we have been using a score and measuring the similarity between a pair of retrieved maps. Empirically, probable maps approximating the particle structures gave a score smaller than a threshold value, but the reasons for the effectiveness of the score are still unclear. In this study, the score is characterized in terms of the phase differences between the structure factors of the retrieved maps, the usefulness of the score in screening the maps retrieved from experimental diffraction patterns is demonstrated, and the effective resolution of similarityscoreselected maps is discussed.1. Introduction
Xray diffraction imaging (XDI) is a technique for visualizing the structures of noncrystalline particles of size in the micro to submicrometre range (Miao et al., 1999, 2015; Nakasako, 2018; Nakasako et al., 2020). In an XDI experiment, the particle to be imaged is irradiated by an Xray beam with almost complete spatial coherence, and the Fraunhofer diffraction pattern of the particle is recorded at a high sampling frequency. When the sampling frequency satisfies the oversampling condition (Miao et al., 2003a), the of the particle projected along the direction of the incident Xray beam is, in principle, reconstructed from the oversampled diffraction amplitudes alone using phaseretrieval (PR) algorithms (Fienup, 1978, 1982).
Owing to the penetration power of shortwavelength Xrays, XDI is advantageous for visualizing whole structures of noncrystalline particles without sectioning and chemical labeling. Therefore, XDI has been applied to structural analyses of noncrystalline particles in material sciences and biology by using synchrotron radiation (SR) Xrays (Williams et al., 2003; Shapiro et al., 2005; Miao et al., 2006; Nishino et al., 2009; Jiang et al., 2010; Takayama & Nakasako, 2012; Nam et al., 2013; Takayama et al., 2018; Kobayashi et al., 2018a) and Xray freeelectron laser (XFEL) pulses (Seibert et al., 2011; Loh et al., 2012; Nakasako et al., 2013; Takahashi et al., 2013; GallagherJones et al., 2014; Hantke et al., 2014; Xu et al., 2014; Kimura et al., 2014; Oroguchi et al., 2015; Takayama et al., 2015a; van der Schot et al., 2015; Ekeberg et al., 2015; Kobayashi et al., 2016a; Kameda et al., 2017; Oroguchi et al., 2018; Nakasako, 2018; Nakasako et al., 2020; Ayyer et al., 2021; Cho et al., 2021; Kobayashi et al., 2021; Uezu et al., 2023). The resolutions of the structure analyses were several tens of nanometres and sometimes reached several nanometres.
When the diffraction pattern of a particle is recorded without any pattern loss or noise, the projected ). However, PR calculations for experimental diffraction patterns frequently give nonrealistic maps due to the loss of smallangle regions hidden by the beamstop and Poisson noise in Xray detection (Huang et al., 2010; Kobayashi et al., 2014; Takayama et al., 2015b; Sekiguchi et al., 2016, 2017). In particular, in our experiences the loss of larger smallangle regions makes the convergence of PR calculations to realistic maps more difficult (Kobayashi et al., 2014), because the smallangle regions contain structural information on the overall shape and the total electrons of the particle.
of the particle can be, in principle, retrieved (Barakat & Newsam, 1984As typical examples, we show PR calculations for singleshot diffraction patterns from clusters of colloidal gold particles in Fig. 1 (Sekiguchi et al., 2017). In the diffraction pattern in Fig. 1(a), the area of the lost smallangle region was more than three times the reciprocal of the cluster size, and the PR calculations yielded realistic maps with approximate probabilities of 50% [Fig. 1(b)]. The diffraction pattern of Fig. 1(c) loses the smallangle region as well as the pattern in Fig. 1(a), and signaltonoise ratios beyond 10 µm^{−1} were smaller than those in Fig. 1(a). As a result, the probability yielding realistic maps was approximately 30% [Fig. 1(d)].
Therefore, to efficiently perform XDI structural analysis, protocols and/or metrics are necessary for screening the retrieved maps from a number of independently performed PR calculations (Chen et al., 2007; Martin et al., 2012; Park et al., 2013; Rodriguez et al., 2013; Kobayashi et al., 2014; van der Schot et al., 2015; Ekeberg et al., 2015; Sekiguchi et al., 2016; FavreNicolin et al., 2020). For screening maps (Sekiguchi et al., 2017), we propose the use of the similarity score (Miao et al., 2003b), defined as
where ρ_{i}(x, y) denotes the projected electron density of the ith map. As XDI theory is formulated based on the first Bornapproximation on a weakphase object, ρ_{i}(x, y) actually represents the projected electron density and not the complex transmission function formulated in ptychography (Pfeiffer, 2018).
Our previous structural analyses on the aggregates of colloidal metal particles (Sekiguchi et al., 2017; Oroguchi et al., 2018) showed that probable images were obtained with a high probability when each pair of the retrieved maps among a set of independently retrieved maps displayed similarity scores smaller than a threshold value (0.2) (Sekiguchi et al., 2017) (Fig. 1). In contrast, when PR calculations failed, the resultant maps were inconsistent with each other and yielded similarity scores greater than the threshold value. The threshold value acts as a necessary condition to extract probable maps, although it is insufficient for the purpose because the score becomes smaller than the threshold value even for pairs of failed maps with similar structures. Owing to the success and usefulness in screening out nonrealistic maps, we have been using the similarity score for XDI structural analyses (Kobayashi et al., 2016a,b; Sekiguchi et al., 2017; Oroguchi et al., 2018; Nakasako et al., 2020; Kobayashi et al., 2021). However, it is unclear as to why the similarity score works as a good metric for screening retrieved maps.
In this study, we investigate how the similarity score works in the screening of retrieved maps in XDI. We describe equation (1) in to understand the correlation between the quality of the retrieved maps, and characterize the similarity score in terms of phase differences between the structure factors of the maps. Through screening the maps retrieved from diffraction patterns obtained in XFELXDI and SRXDI experiments, we assess the usefulness of the score and the resolutiondependent contributions of phase differences to the score.
2. Theoretical background
2.1. Interpretation of the similarity score in reciprocal space
2.1.1. Expression of the similarity score by Fourier transform
To characterize the similarity score in ith projection map, ρ_{i}(x, y) in equation (1), as the inverse Fourier transform of the F_{i}(S), where S is the scattering vector. Equation (1) in can be written as
we express theAccording to the PR algorithm (Fienup, 1982), the final map is given by the inverse Fourier transform of the composed of the observed F_{obs}(S), and the phase in the final calculation cycle, α_{i}(S). Therefore, equation (2) is modified as
Equation (3) indicates that the similarity score depends on the phase differences between the structure factors of the two maps.
We designate the term as the `lack of identity' (LoI) [Fig. 2(a)], analogous to the lack of closure of structure factors in the multiple method (Blow & Crick, 1959). Using the following relation on the difference and sum of the phase terms [Fig. 2(b)],
then equation (3) is modified as
The trigonometric terms of Δα_{ij}(S) originating from the LoI act as weights on the A(S), which is a product of F_{obs}(S) and the phase averaged between the structure factors of the ith and jth maps. As the structure amplitudes at the lower resolution are greater than those at the higher resolution, the phase differences of the structure factors at the lower resolution dominate the score. With respect to two similar maps, the phase difference of the structure factors at low resolution will be as small as zero and reduce the numerator.
2.1.2. Resolutiondependent variation of the Fourier term
As described above, the trigonometric terms of Δα_{ij}(S) and the structure amplitudes vary depending on the resolution. This implies that the similarity score can display a correlation with the resolution of the pair of maps, although the similarity score does not obviously depend on the resolution. Thus, it is meaningful if we modify each of the denominator and numerator in equation (5) into formulae suitable for monitoring the resolutiondependent variations of the trigonometric term. By using , the numerator and denominator in equation (5) satisfy the following inequalities,
Then, using the equation f(x)g(x) = f(x)g(x), the righthandside term in each of the above inequalities is expressed as the sum of the integrals on the Ndivided resolution shells and each integrand is separated as
Therefore, the separated trigonometric term may be helpful for numerically evaluating the influence of each resolutionshell to the similarity score.
2.1.3. Influences of phase differences in the resolution shell
Based on equation (6), to roughly estimate the influences of the phase differences in each resolution shell we use the following two quantities averaged over the maps in the narrow kth resolution shell,
where P(Δα_{k}) is the normalized appearance frequency of Δα_{k}. Hereafter, we designate the first and second equations as averaged cosine and sine terms in the kth shell, respectively.
Using equation (7), the influence of the trigonometric terms is qualitatively evaluated. For maps similar to each other, small phase differences between the structure factors in the lowresolution shell will yield values of the averaged cosine and sine term as large as 1 and 0, respectively. In contrast, for highresolution shells, where the ambiguity of structure amplitudes due to Poisson noise causes large phase differences between the two structure factors, the averaged cosine term decreases, while the averaged sine term increases.
2.2. Reference values from special cases
As equation (6) is complicated, here we show values of the averaged cosine and sine terms in two special cases, which may be useful as references for roughly estimating the influences of the phase differences on the similarity score.
In the first special case, we assume that a particle has a circular shape with radius r_{0} and uniform density. Then, the is written by the following equation (Born & Wolf, 2001),
where J_{1} is the firstorder cylindrical Bessel function and C is a constant. The phase value between any pair of adjacent zerocrossing points of J_{1} is zero or π. When PR calculations for the diffraction amplitudes give two maps of circular shapes with different radius, the phase difference values are zero or π in any narrow resolution shell. Therefore, the phase differences between the two maps in the kth resolution shell are a constant.
In such a case, without the two inequalities introduced for the general case described above, the similarity score in the kth resolution shell is simplified as
where Δα_{k} is a constant phasedifference value in the kth resolution shell. When Δα_{k} = 0, the sine and cosine terms are 0 and 1, respectively, resulting in the similarity score in the resolution shell becoming 0. In contrast, when Δα_{k} = π, the maximum phase difference, then the sine and cosine terms are 1 and 0, and the score diverges to infinity.
In this regard, we express an electron density distribution using the multipole expansion (Stuhrmann & Miller, 1978). Then, the electron density distribution is the sum of the first monopole term, the second dipole term, the third quadrupole and higherorder terms. If the weight for the first monopole term is approximated as uniform, the has the formula for a circular shape described above. In addition, when the monopole term approximating the size of the density distribution has a large electron density contrast (Ibel & Stuhrmann, 1975), the Fourier transform of the monopole term is dominant in the of the density distribution in the smallangle region. Therefore, this simple numerical calculation may be helpful to roughly evaluate the influences of the phasedifference terms in equation (6) on the similarity score.
In the next special case, we numerically calculate the averaged cosine and sine terms for the case that the kth resolution shell. Then, the values of the averaged cosine and sine terms approach 0.64 as
of phase differences is random among the structure factors of retrieved maps in theHereafter, this value is designated the random phase limit.
2.3. Figure of merit for estimated phase
In Xray crystallography and ), based on the probability distribution of phase values in the experimental estimation, as follows,
(TEM) of biological macromolecules, the reliability of a reconstructed map is evaluated using the figure of merit (FoM) (Blow & Crick, 1959where P[α_{k}(S)] is the probability of the kth bin of the phase values at the scattering vector S. The reconstructed maps displaying a FoM greater than the threshold of 0.5 are regarded as interpretable (Lunin & Woolfson, 1993; Perrakis et al., 1997). The threshold indicates that the phase values are distributed within ±60° from the true value (Drenth, 1994). The FoM for the structure factors of two maps is in correlation with the LoI as
Thus, the similarityscoreselected maps were evaluated using FoM and LoI.
2.4. Phaseretrieval transfer function for estimated phase
In XDI structural analyses, the phase retrieval transfer function (PRTF) is frequently used to estimate the effective resolution for a set of retrieved maps (Chapman et al., 2006). The PRTF for a set of retrieved maps is defined as
where M is the number of retrieved maps incorporated into the calculation, and 〈…〉_{radial} refers to circular averaging with respect to the scattering vector length. The effective resolution is determined from the reciprocal of the scattering vector length, where the PRTF curve decreased to a specified threshold value. The PRTF for the structure factors of the two retrieved maps is derived using equation (4) as
where we use the fact that the cosine term is positive for −π ≤ Δα_{ij}(S) ≤ +π. The effective resolution estimated using the PRTF will be discussed in conjunction with the similarity score in the Discussion section.
3. Experimental procedure and structure analysis
3.1. Specimen preparation
For XFELXDI experiments, colloidal gold particles with a mean diameter of 250 nm (BBI Solutions, UK) were dispersed on custommade specimen disks, each of which had a 1 mm × 1 mm Si_{3}N_{4} membrane window of 100 nm thickness (Norcada, Canada). Further details were described in our previous publications (Sekiguchi et al., 2014a, 2016; Kobayashi et al., 2016a,b; Nakasako, 2018; Nakasako et al., 2020).
For the SRXDI experiments, we used commercially available specimen disks with a 5 mm × 5 mm Si_{3}N_{4} membrane window (Silson Ltd, UK). The windows were coated with ∼15 nmthick carbon layers using a JEE420 vacuum evaporator (Jeol, Japan) and then covered with poly Llysine layers (SigmaAldrich, USA) to assist the adhesion of the specimen particles to the Si_{3}N_{4} membranes (Takayama & Yonekura, 2016; Kobayashi et al., 2016b). We prepared specimen disks adsorbing colloidal gold particles with a mean diameter of 400 nm (SigmaAldrich, USA) or gold urchin particles synthesized according to the literature (Wang et al., 2012).
3.2. XFELXDI experiments and data processing
XFELXDI experiments were performed using our custommade apparatus KOTOBUKI1 (Nakasako et al., 2013) at beamline BL3 of SACLA (Tono et al., 2013). Focused XFEL pulses with a photon energy of approximately 5.5 keV (corresponding to a wavelength of 0.225 nm) had intensities of 10^{10}–10^{11} photons pulse^{−1} µm^{−2} and almost complete spatial coherence (Kobayashi et al., 2018a). The specimen mounted on the goniometer of KOTOBUKI1 was scanned with a 50 µm step against the incident XFEL pulses, and singleshot diffraction patterns were recorded using multiport CCD (MPCCD) octal and dual detectors (Kameshima et al., 2014). After each scan, the recorded diffraction patterns were processed using the GSITENNO program suite (Sekiguchi et al., 2014a,b), which extracted diffraction patterns from a cluster of colloidal gold particles with a signaltonoise ratio greater than 2 at a resolution of 15 µm^{−1}. Further details of the experimental setup, procedures and data processing were reported previously (Sekiguchi et al., 2014a, 2016; Kobayashi et al., 2016a,b; Nakasako, 2018; Nakasako et al., 2020; Kobayashi et al., 2021; Uezu et al., 2023).
3.3. XDI experiments using synchrotron Xrays and data processing
The SRXDI experiments were performed using an originally developed atmospheric XDI system (Takayama et al., 2018, 2021) installed at the imaging station of BL24XU in SPring8 (Takayama et al., 2020). Specimen particles were irradiated with the nearly planewave region of the incident Xrays produced using a 30 µm pinhole, which was placed approximately 1.5 m upstream of the specimen position.
An aggregate of gold urchin particles was irradiated by Xrays with a photon energy of 8.000 keV (corresponding to a wavelength of 0.1550 nm), and the diffraction pattern was recorded using a PILATUS 100k detector (Dectris Ltd, Switzerland) (Kraft et al., 2009) placed approximately 4.1 m downstream of the specimen. An aggregate of 400 nm colloidal gold particles was irradiated by Xrays with a photon energy of 8.310 keV (corresponding to a wavelength of 0.1492 nm), and the diffraction pattern was recorded using an EIGER X 1M detector (Dectris Ltd, Switzerland) (Dinapoli et al., 2011) placed approximately 3.2 m downstream of the specimen.
For each specimen, a background pattern was recorded from an area of the Si_{3}N_{4} membrane yielding no speckles, and was subtracted from the diffraction pattern of the specimen. Centrosymmetric averaging was applied to the backgroundsubtracted diffraction pattern to improve the signaltonoise ratio owing to good Friedel symmetry.
In the SRXDI experiments described above, as the coherent et al., 2018). In this case, the background scattering predominantly came from the Si_{3}N_{4} membrane irradiated by the foot region of the pinholediffracted Xrays, where the spatial coherence was low. The electron density in the membrane was assumed to be uniform at the maximum resolution of the experiment. In addition, the background scattering from the membrane was assumed to be incoherent and the interference between the background scattering and the diffraction wave from the specimen was negligible. Therefore, the background subtraction procedure was routinely applied as reported for SRXDI experiments at BL24XU (Takayama et al., 2018) and BL29XUL (Nishino et al., 2009; Nam et al., 2013; Kobayashi et al., 2018b) at SPring8 and also for XDI experiments at BL9C at Pohang Light Source II (Ahn et al., 2021).
of SR Xrays from the undulator of BL24XU was around 0.1% due to the size of the light source (electron bunch) in the SPring8 storage ring (approximately 300 µm along the horizontal direction), we extracted the coherent part of the incident Xrays using a pinhole with a diameter of 30 µm, and a specimen was irradiated by the planewave part at the of the diffraction from the pinhole (Takayama3.4. PR calculation and multivariate analysis
We performed PR calculations for the diffraction patterns using a custommade Python3 software, in which the hybridinput–output (HIO) (Fienup, 1982), shrinkwrap (SW) (Marchesini et al., 2003) and error reduction (ER) algorithms (Gerchberg & Saxton, 1972) were implemented. Each of 1000 independent PR calculations started with a randomvalued map different among the 1000 calculations. An initial support was set as the area where the magnitude of the realspace autocorrelation function calculated from the diffraction pattern was greater than 4% of the maximum. Each projected was retrieved by 10000 iterations of the HIO cycles with SW modifications at every 100 HIOcycles and additional 1000 iterations of the ER calculation. Details of HIO and SW were reported previously (Kodama & Nakasako, 2011; Oroguchi & Nakasako, 2013).
Based on the assumption that realistic maps were obtained at high frequency and similar to each other when a PR calculation is successful, we searched a reference map for a calculation of the similarity score among the set of retrieved maps as follows. The 1000 retrieved maps were first translated to coincide the center of gravity of the density with that of an arbitrarily selected reference map, and we calculated the similarity score for whole combinations of 1000 maps (499500 pairs). Then, we selected the pair of maps yielding the smallest similarity score as the most probable pair, and one of the pair was used as a reference map for the subsequent calculation. Each map was translated to maximize phaseonly correlation (Kuglin & Hines, 1975) against the reference map. In the PR calculation, several maps in the πrotation relative to the reference map appeared owing to the centrosymmetry of the diffraction patterns. These maps were corrected with respect to πrotation during translational alignment using the phaseonly correlation. After the correction, the similarity scores for whole combinations of 1000 maps were again calculated against the reference. It should be noted that either of the map pair gave almost the same result of the calculated similarity scores.
Independently, the maps were classified into ten groups by Kmeans clustering (MacQueen, 1967) without the information from the similarity score. When assuming that the maps are classified into ten groups, we minimize the sum of squared distances between the maps and the centroids of the groups defined as
where y_{im} is a vector indicating the position of the ith map belonging to the mth group and 〈y_{m}〉 is the centroid of the mth group. For visualizing the distributions of the classified ten groups, the map positions were projected onto a plane spanned by the two principal component (PC) vectors, which had the largest and second largest eigenvalues in the principal component analysis (PCA) (Jolliffe & Cadima, 2016) applied to the 1000 retrieved maps. It should be noted that, on the plane spanned by the first and second PC vectors, clusters separated along the thirdorder or higherorder vectors will be overlayed because the Kmeans clustering was applied before the dimensionality reduction by PCA. Details of the PCA for a number of PR maps were reported previously (Sekiguchi et al., 2016). The scikitlearn library (Pedregosa et al., 2011) was used for the Kmeans clustering and PCA.
Finally, in each class, the phase distributions among the structure factors of the maps were evaluated using the FoM at every scattering vector [equation (9)] (Sekiguchi et al., 2017), and the phase differences in the structure factors between any pair of maps were represented using the averaged cosine and sine terms of equation (7) in each resolution shell.
4. Results
Here, we performed the screening of maps phaseretrieved from experimental diffraction patterns using the similarity score in real space and interpreted the score in terms of phase differences between the structure factors of the retrieved maps. In the analyses, we targeted diffraction patterns from aggregates of colloidal gold particles, because the known size and shape of the particles made identifying realistic maps easy.
4.1. Application to singleshot diffraction patterns recorded using XFEL pulses
Here, we assessed the similarity score for the maps retrieved from singleshot diffraction patterns recorded using XFEL pulses (Fig. 1). As a lot of singleshot diffraction patterns can be collected within a short time in our XFELXDI experiments (Kobayashi et al., 2016a; Nakasako, 2018), efficient and automatic screening of probable and nonrealistic maps is necessary in the structure analyses.
One thousand maps were retrieved from the diffraction pattern of an aggregate of ten colloidal gold particles in Fig. 1(a), and were divided into ten classes on the plane spanned by the first and second PC vectors [Fig. 3(a)]. The maps in classes 2, 5, 8 and 9 forming a cluster on the plane displayed similarity scores smaller than 0.19 against the reference map, which was one of the pair yielding the smallest similarity score among all the maps [Fig. 3(b)].
Here, we define a quantity in
designated the Fourier error of a retrieved map aswhere F_{cal}(S_{x}, S_{y}) is the of a retrieved map masked by the support. Because this quantity is related to the target function of the phase retrieval to be reduced, we examined whether this quantity for each map was useful for extracting realistic maps. In this case, as the frequency distributions of the Fouriererror values of the ten classes were heavily overlapped [Fig. 3(c)], we abandoned the use of this quantity for extracting maps of classes 2, 5, 8 and 9 [see Fig. 3(d)].
Fig. 3(d) depicts the map in each of the ten classes, yielding the smallest similarity score against the reference map. In the maps of classes 2, 5, 8 and 9, the ten colloidal gold particles appeared with clear borders and were separated from each other, indicating that the maps in the four classes were successfully retrieved. Among the four correct classes, the maps of class 8 yielded the smallest similarity scores [Fig. 3(b)]. The PR calculations for the other six classes failed, as judged by the unclear and blurred electron densities of the particles. These results imply that the similarity score is a good measure for screening maps in real space. As indicated by the similarity scores, the probable maps were similar to each other, whereas nonrealistic maps were different from each other [Fig. 3(b)].
The similarity of the maps in real space was evaluated using the FoM [Fig. 3(e)] and LoI [Fig. 3(f)] in In the four successfully retrieved classes, particularly classes 2 and 8, the reciprocal regions with FoM values greater than 0.5 extended to a resolution of 18 µm^{−1}. In contrast, the FoM values of the nonrealistic classes were smaller than 0.5, with a resolution lower than 10 µm^{−1}.
Fig. 3(f) shows the resolutiondependent variations of the cosine and sine terms of equation (7). For the pair of maps yielding the smallest similarity score among the 1000 phaseretrieved maps, the phase differences gradually approached the random phase limit [equation (8)] beyond a resolution of 16 µm^{−1}. In the realistic classes, the phase differences likely varied in the range 20–30° among the maps up to a resolution of 8 µm^{−1}. In the nonrealistic classes, judging from the values close to the random phase limit of 0.64 [see equation (8)], the phase differences became random even at a resolution of approximately 2 µm^{−1}; this might be a major cause of blurry densities and unclear edges of the averaged maps. It should be noted that there is a boundary of the datamissing area due to the beamstop around the resolution of approximately 2 µm^{−1} [Fig. 1(a)].
Next, we assessed the similarity score for the maps retrieved from another singleshot diffraction pattern shown in Fig. 1(c). Among the ten classes of the 1000 retrieved maps [Fig. 4(a)], only the maps in class 3 showed similarity scores smaller than 0.18 against the reference map [Fig. 4(b)]. In this case, the of the Fouriererror values for maps in class 3 had a peak at 0.026 [Fig. 4(c)], but, in contrast to the similarity score [Fig. 4(b)], the overlap of the frequency distributions among the ten classes made it difficult to extract most of the maps in class 3. Fig. 4(d) depicts the map in each of the ten classes displaying the smallest similarity score against the reference. In the class 3 map [Fig. 4(d)], the three major and three minor densities of colloidal particles were clear, and the minor density at the upper edge was interpreted as the particle being located in the foot part of the XFEL pulse. In the maps of classes 1, 7 and 9, which yielded similarity scores greater than 0.2, the edges of the three major particles were unclear and blurred, indicating failures of the PR calculations. Regarding the FoM of the structure factors of the maps in each class [Fig. 4(e)], classes 3 and 8 maintained FoM values greater than 0.5 beyond a resolution of 10 µm^{−1}. In classes 2, 5, 6 and 10, the FoM values were directiondependent, suggesting that the maps in the four classes had anisotropic similarity.
As shown in Fig. 4(f), the phase differences between the pair of maps yielding the best similarity score approached the random phase limit [equation (8)] beyond a resolution of 14 µm^{−1}. In contrast, the maps in each of classes 2, 4, 5, 6, 8 and 10 displayed sudden changes in the cosine and sine terms in the resolution range 1–2 µm^{−1}, around the boundary of the datamissing area due to the beamstop [Fig. 1(c)]. The two terms for the maps in each of classes 1, 7 and 9 indicated greater phase differences of structure factors up to a resolution of 17 µm^{−1}.
Based on the results, we concluded that the similarity score is good for screening maps phaseretrieved from singleshot diffraction patterns. The similarity score in real space was more convenient for automatically screening maps than the FoM and the two trigonometric terms in ]. The trigonometric terms in are important to explain how the similarity score works.
[equation (7)4.2. Application to diffraction patterns recorded using synchrotron Xrays
In this section, we apply the similarity score to the maps retrieved from the diffraction patterns of clusters composed of several colloidal gold particles recorded in the SRXDI experiments. In the SRXDI experiment, the collection of diffraction patterns from specimens with sizes of several micrometres is possible owing to the wider spatially coherent irradiation area of approximately 10 µm diameter and the smaller beamstop covering a much smaller diffraction angle than those in XFELXDI (Takayama et al., 2018; Kobayashi et al., 2018b). In addition, good signaltonoise ratios in highdiffractionangle regions might be expected owing to both the relatively large scattering of specimens with sizes of several micrometres and the exposure time.
In Fig. 5, we summarize the application of the similarity score for the maps retrieved from the diffraction pattern of an aggregate composed of more than 20 colloidal gold particles of 400 nm diameter. Of the ten classes divided from the 1000 retrieved maps, only the maps of class 2 were localized in the plane spanned by the two PC vectors [Fig. 5(b)]. Fig. 5(b) depicts the map in each of the ten classes, yielding the smallest similarity score against the reference map. Only the maps of class 2 were realistic, as judged from the clear borders and voidlike fine structures of the particles, and yielded a similarity score smaller than 0.08 against the reference [Fig. 5(c)]. As the Fouriererror values of the maps in class 2 displayed a single peak in the [Fig. 5(d)] and correlated with the similarity score values, the Fourier error may be used to extract the class 2 maps in this case.
Fig. 5(e) shows the resolutiondependent variations of the averaged cosine and sine terms of equation (7). For a pair of maps yielding the smallest similarity score, the phase differences in the structure factors of the maps gradually approached the random phase limit beyond a resolution of 30 µm^{−1} [equation (8)]. The two terms of class 2 also asymptotically reached this limit beyond 30 µm^{−1}, while those of the other nonrealistic classes reached the value within 5 µm^{−1}.
As a second example, we assessed the similarity score for maps retrieved from the diffraction pattern of an aggregate of gold urchin particles with irregular shapes (Fig. 6). As the diffraction pattern included a part of the central speckle in the very small angle region and was recorded with good signaltonoise ratios up to a resolution of 25 µm^{−1} [Fig. 6(a)], almost all of the maps were successfully retrieved [Fig. 6(b)]. Although the correct maps displayed similarity scores smaller than 0.19, the maps were divided into classes 2–10, probably because of the small differences in the local fine structures, such as convections and protrusions on the surfaces [Figs. 6(b) and 6(c)]. All of the class 2 maps yielded similarity scores smaller than 0.09 against the reference. The Fourier error values of all maps were approximately 0.0241. As the values of the nonrealistic maps of class 1 overlapped with the frequency distributions of the other class, it was difficult to exclude the Fourier errors in the maps of class 1 [Fig. 6(d)].
In Fig. 6(e), for the two maps yielding the best similarity score, the phase differences of the structure factors were still nonrandom beyond a resolution of 25 µm^{−1}. For only the maps of class 2, the resolutiondependent variations of the cosine and sine terms of equation (7) were similar to those of the best pair and displayed no sudden changes up to a resolution of 5 µm^{−1}. Beyond 23 µm^{−1}, the two terms reached the random phase limit of 0.64 [equation (8)], probably because the phase differences among the maps became random.
From the structural analyses of the two diffraction patterns in the SRXDI experiments, the similarity score in real space acts as a good measure in screening retrieved maps, and the trigonometric terms of equation (7) in were helpful for understanding the correlation between the similarity score and the validity of the maps.
5. Discussion
In this study, we characterized the similarity score in correlation with the phase differences in the structure factors of the retrieved maps. Through application to structural analyses for experimental diffraction patterns, we demonstrated the usefulness of the similarity score for screening phaseretrieved maps. Here, we discuss the weak point in the use of the similarity score, the estimation of the effective resolution of the retrieved maps using the score, and the utilization of the score for efficiently performing PR calculations.
5.1. Selection of maps using the similarity score
In this study, for selecting realistic maps, which approximate the shape, size and internal structure of a particle, we use the similarity score under the assumption that realistic maps yield the smallest similarity score among the retrieved maps. However, when pairs of nonrealistic maps different from the structures of particles are quite similar, this assumption cannot be made. In our experience in structure analyses of aggregates of colloidal gold particles, the breakdown of the assumption occurs at a probability smaller than 0.1%. However, crossvalidation using known structural information, for instance from et al., 2014; Takayama et al., 2015a; Oroguchi et al., 2018; Kobayashi et al., 2021).
(EM) and fluorescence microscopy, is still necessary for the selected maps, as carried out in our previous works (KobayashiIn this regard, we describe two examples on crossvalidation in the et al., 2000).
of protein molecules. Even in the estimation of phase set using the multiple heavyatom replacement method in protein crystallography, we validated the heavyatom positions in the of derivative crystals using the phase set in the centric zone of the Bragg reflections obtained from cryogenic TEM (ToyoshimaIn smallangle Xray scattering for proteins in solution, the molecular shapes of protein molecules are predicted using an ab initio algorithm (Svergun et al., 2001). As the algorithm is applied to a onedimensional scattering profile under restraints with respect to the arrangement of small spheres approximating the molecular shape, both realistic and nonrealistic models appear in more than 500 trial calculations. Then, after classification of the models using multivariate analysis (Oide et al., 2018), crossvalidation using structural information from other imaging techniques is indispensable for finding realistic molecular structures (Oide et al., 2020; Oide & Nakasako, 2021).
The similarity score may be a useful tool for screening realistic maps from a number of trial calculations. However, as we have no experimentally determined phase data in XDI, we keep in mind the possibility of the breakdown of the assumption. Therefore, even for maps yielding small similarity scores, validation for the shape, size and internal structure of the particle is always necessary.
5.2. Similarity score and differences in structure factors of maps at low resolution
In the four examples of structure analyses (Figs. 3–6), the frequency distributions of the similarity scores of phaseretrieved maps in each class are likely correlated with the resolutiondependent variations of the cosine and sine terms for the phase differences of the structure factors between the maps. When the similarity scores in a class are distributed in a range smaller than 0.2, the cosine term is close to 1 and the sine term is near 0 in the smallangle region [Figs. 3(f), 4(f), 5(e) and 6(e)]. Owing to the structure amplitudes of the smallangle region, which are much greater than those of the highangle region [Figs. 1(a), 1(c), 5(a) and 6(a)], when the phase differences of the structure factors of maps are small in the smallangle region, the contribution of the structure factors to the numerator of equation (6) is reduced. Therefore, the small value of the similarity score is interpreted as the retrieved maps viewed at low resolution being similar to each other.
As the contribution of fine structures to the structure factors are dominant at high resolution, the screening of retrieved maps using the similarity score predominantly focuses on the lowresolution structures of the retrieved maps. At low resolution, the FoM takes values close to 1 [Figs. 3(e) and 4(e)], as expected from equation (9). Therefore, the similarity score may be useful in the preliminary screening of realistic maps before applying, for instance, the unweighted pair group method with arithmetic mean analysis, which is the distancematrix method and employs sequential clustering to yield a rooted phylogenetic tree of image data as used in XFELXDI structure analysis (Ekeberg et al., 2015).
5.3. Effective resolution estimated using the similarity score
As the similarity score is dominated by the phase differences between structure factors of retrieved maps in the smallangle region, we discuss how the effective resolution for a set of retrieved maps is quantitatively estimated using the resolutiondependent variation of the cosine (or sine) term in equation (7). Then, we referred to the fine structures of the averaged maps, such as resolved narrow gaps between colloidal particles (arrows in the averaged map in Fig. 7 and Table 1). As the phase differences in a resolution shell between two maps are random, the cosine (or sine) term approaches the random phase limit (0.64) [equation (8), Figs. 3(f), 4(f), 5(e) and 6(e)], we adopted the reciprocal of the scattering vector length, where the cosine curve for sets of retrieved maps approaches the randomphase limit, as an effective resolution of the averaged map.

As a result, the effective resolution estimated in each of Figs. 7(b) and 7(c) was approximately twice the gap dimensions. From the point of view of the resolution in microscopy that is given by the Fourier component with the shortest period, half of the minimum period of the Fourier component is resolved. In Figs. 7(a) and 7(d), as the cosine terms were still greater than 0.64 at the highest resolution recorded, the effective resolution of the averaged map was assumed to be the highest resolution of the diffraction pattern used in the PR calculation (Table 1).
We also estimated the effective resolution using the PRTF curves [equation (10)]. The curves were calculated independently for ten and 50 maps from the best classes in each structure analysis and additionally for two maps yielding the smallest similarity score as a reference (Fig. 7). For maps displaying small variations in electron density, the greater number of maps used in the PRTF calculation likely correlated with the more rapid decrease of the curves [Figs. 7(a)–7(c)]. In contrast, for the maps that were very similar to each other, the PRTF curves were almost independent of the number of maps [Fig. 7(d)]. A comparison of the four cases indicated that the PRTF curves were likely advantageous for monitoring the degree of the convergence of the maps.
When the threshold for estimating the effective resolution of the maps from the PRTF curve was set to 0.5 or 1/e (∼0.37) (Chapman et al., 2006), the estimated effective resolution was lower than that to resolve the narrow gaps between particles. Instead, we tried to use the halfbit information threshold curve (van Heel & Schatz, 2005), which was proposed in conjunction with the Fourier shell correlation (FSC) curve (Rosenthal & Henderson, 2003) for estimating the effective resolution in single particle analysis using TEM. Although the definition of the FSC curve is different from that of the PRTF curve, both curves essentially measure the distribution of phase values. The halfbit curve, which is correlated with the information content in voxels (pixels) in a resolution shell, is known to allow a reliable interpretation of the resolution level in TEM analysis (van Heel & Schatz, 2005). The effective resolution estimated from the scattering vector length, where the PRTF curve crossed the halfbit curve, was comparable with the doubled gap dimension between colloidal particles (Fig. 7 and Table 1).
As a result, the effective resolution estimated using the cosine term and the randomphase limit of 0.64 was comparable with that using the PRTF and halfbit curves. The estimation of the effective resolution of the retrieved maps is still under debate – further experiences regarding the structure analyses of various aggregates of metal particles are necessary.
5.4. Similarity score and landscape on the distribution of retrieved maps
In the previous study, we assumed that the phaseretrieved maps were distributed on a funnelshaped landscape in multidimensional space spanned by the similarity score and pixel values of the maps, reflecting the structure, as schematically illustrated in Fig. 8(a) (Sekiguchi et al., 2017). Based on this idea, the retrieved maps in Figs. 3–6 were plotted in the threedimensional space spanned by the similarity score and the two PC vectors representing the major characteristics of the electron density distributions in all the retrieved maps [Figs. 8(b)–8(e)]. The maps were roughly distributed along funnelshaped surfaces, and the distributions were different from each other probably due to the area of the missing smallangle regions, the signaltonoise ratios and the oversampling ratios. The maps displaying small similarity scores were concentrated near the bottom part of the funnel. In contrast, the other maps with large scores were distributed around the upper edges of the funnel [Fig. 8(b)], or in local minima [Fig. 8(c)].
This representation is helpful for understanding characteristics of the funnelshaped surface. As even maps displaying large scores are located in a local area on the twodimensional planes, the landscape of the area is expected to be rugged as illustrated in Fig. 8(a). In the case of the retrieved maps of the aggregate of gold urchin particles, the nine classes discretely distributed near the bottom of the funnel [Fig. 8(e)] suggested local and shallow minima even at the bottom of the funnel, probably because of the small differences in the fine structures.
The visualized distributions encouraged us to steer the PR calculations toward the most probable solution for the observed diffraction pattern by referring to the similarity scores during the progress of the calculations. For instance, in parallelly conducted PR calculations, the modification of maps by those near the bottom of the funnel may increase the probability of successful calculations (Yoshida et al., 2024).
Acknowledgements
We thank Dr Kensuke Tono and the SACLA engineering team of the Japan Synchrotron Radiation Research Center for their help in the XFELXDI experiments, and Professor Yasushi Kagoshima and the members of his laboratory for their help in the SRXDI experiments. SRXDI diffraction data were collected at BL24XU of SPring8 (Proposal No. 2016B3201, 2018A3201). Data processing for XFELXDI was performed on an HPC computer at the SACLA facility.
Funding information
This study was supported by grants for XFEL key technology and the Xray
Priority Strategy Program to MN from The Ministry of Education, Culture, Sports, Science and Technology (MEXT). In addition, this study was supported by grants from The Japan Society for The Promotion of Science to MN (No. jp17654084, jp23120525, jp24654140, jp25120725, jp16H02218 and jp21H01050) and YT (No. jp16K21621 and jp19K14678), grants from MEXT to MN (No. jp15076210, and jp20050030) and by The CREST grant from the Japan Science and Technology Agency (No. JPMJCR2233 to YT). XFELXDI diffraction data were collected at SACLA (proposals 2013B8049 and 2014A8033)References
Ahn, K., Cho, I. H., Kim, J., Lee, S. Y., Sung, D., Jung, C., Song, C., Kang, H. C. & Noh, D. Y. (2021). J. Synchrotron Rad. 28, 505–511. Web of Science CrossRef CAS IUCr Journals Google Scholar
Ayyer, K., Xavier, P. L., Bielecki, J., Shen, Z., Daurer, B. J., Samanta, A. K., Awel, S., Bean, R., Barty, A., Bergemann, M., Ekeberg, T., Estillore, A. D., Fangohr, H., Giewekemeyer, K., Hunter, M. S., Karnevskiy, M., Kirian, R. A., Kirkwood, H., Kim, Y., Koliyadu, J., Lange, H., Letrun, R., Lübke, J., Michelat, T., Morgan, A. J., Roth, N., Sato, T., Sikorski, M., Schulz, F., Spence, J. C. H., Vagovic, P., Wollweber, T., Worbs, L., Yefanov, O., Zhuang, Y., Maia, F. R. N. C., Horke, D. A., Küpper, J., Loh, N. D., Mancuso, A. P. & Chapman, H. N. (2021). Optica, 8, 15–23. Web of Science CrossRef Google Scholar
Barakat, R. & Newsam, G. (1984). J. Math. Phys. 25, 3190–3193. CrossRef Web of Science Google Scholar
Blow, D. M. & Crick, F. H. C. (1959). Acta Cryst. 12, 794–802. CrossRef CAS IUCr Journals Web of Science Google Scholar
Born, M. & Wolf, E. (2001). Principles of Optics, 7th (expanded) ed. Cambridge University Press. Google Scholar
Chapman, H. N., Barty, A., Marchesini, S., Noy, A., HauRiege, S. P., Cui, C., Howells, M. R., Rosen, R., He, H., Spence, J. C. H., Weierstall, U., Beetz, T., Jacobsen, C. & Shapiro, D. (2006). J. Opt. Soc. Am. A, 23, 1179–1200. Web of Science CrossRef Google Scholar
Chen, C.C., Miao, J., Wang, C. W. & Lee, T. K. (2007). Phys. Rev. B, 76, 064113. Web of Science CrossRef Google Scholar
Cho, D. H., Shen, Z., Ihm, Y., Wi, D. H., Jung, C., Nam, D., Kim, S., Park, S.Y., Kim, K. S., Sung, D., Lee, H., Shin, J.Y., Hwang, J., Lee, S. Y., Lee, S. Y., Han, S. W., Noh, D. Y., Loh, N. D. & Song, C. (2021). ACS Nano, 15, 4066–4076. Web of Science CrossRef CAS PubMed Google Scholar
Dinapoli, R., Bergamaschi, A., Henrich, B., Horisberger, R., Johnson, I., Mozzanica, A., Schmid, E., Schmitt, B., Schreiber, A., Shi, X. & Theidel, G. (2011). Nucl. Instrum. Methods Phys. Res. A, 650, 79–83. Web of Science CrossRef CAS Google Scholar
Drenth, J. (1994). Principles of Protein Xray Crystallography. Berlin, Heidelberg: SpringerVerlag. Google Scholar
Ekeberg, T., Svenda, M., Abergel, C., Maia, F. R. N. C., Seltzer, V., Claverie, J.M., Hantke, M., Jönsson, O., Nettelblad, C., van der Schot, G., Liang, M., DePonte, D. P., Barty, A., Seibert, M. M., Iwan, B., Andersson, I., Loh, N. D., Martin, A. V., Chapman, H., Bostedt, C., Bozek, J. D., Ferguson, K. R., Krzywinski, J., Epp, S. W., Rolles, D., Rudenko, A., Hartmann, R., Kimmel, N. & Hajdu, J. (2015). Phys. Rev. Lett. 114, 098102. Web of Science CrossRef PubMed Google Scholar
FavreNicolin, V., Leake, S. & Chushkin, Y. (2020). Sci. Rep. 10, 2664. Web of Science PubMed Google Scholar
Fienup, J. R. (1978). Opt. Lett. 3, 27–29. CrossRef PubMed CAS Web of Science Google Scholar
Fienup, J. R. (1982). Appl. Opt. 21, 2758–2769. CrossRef CAS PubMed Web of Science Google Scholar
GallagherJones, M., Bessho, Y., Kim, S., Park, J., Kim, S., Nam, D., Kim, C., Kim, Y., Noh, D. Y., Miyashita, O., Tama, F., Joti, Y., Kameshima, T., Hatsui, T., Tono, K., Kohmura, Y., Yabashi, M., Hasnain, S. S., Ishikawa, T. & Song, C. (2014). Nat. Commun. 5, 3798. Web of Science PubMed Google Scholar
Gerchberg, R. W. & Saxton, W. O. (1972). Optik, 35, 237–246. Google Scholar
Hantke, M. F., Hasse, D., Maia, F. R. N. C., Ekeberg, T., John, K., Svenda, M., Loh, N. D., Martin, A. V., Timneanu, N., Larsson, D. S. D., van der Schot, G., Carlsson, G. H., Ingelman, M., Andreasson, J., Westphal, D., Liang, M., Stellato, F., DePonte, D. P., Hartmann, R., Kimmel, N., Kirian, R. A., Seibert, M. M., Mühlig, K., Schorb, S., Ferguson, K., Bostedt, C., Carron, S., Bozek, J. D., Rolles, D., Rudenko, A., Epp, S., Chapman, H. N., Barty, A., Hajdu, J. & Andersson, I. (2014). Nat. Photon. 8, 943–949. Web of Science CrossRef CAS Google Scholar
Heel, M. van & Schatz, M. (2005). J. Struct. Biol. 151, 250–262. Web of Science PubMed Google Scholar
Huang, X., Nelson, J., Steinbrener, J., Kirz, J., Turner, J. J. & Jacobsen, C. (2010). Opt. Express, 18, 26441–26449. Web of Science CrossRef PubMed Google Scholar
Ibel, K. & Stuhrmann, H. B. (1975). J. Mol. Biol. 93, 255–265. CrossRef PubMed CAS Web of Science Google Scholar
Jiang, H., Song, C., Chen, C.C., Xu, R., Raines, K. S., Fahimian, B. P., Lu, C.H., Lee, T.K., Nakashima, A., Urano, J., Ishikawa, T., Tamanoi, F. & Miao, J. (2010). Proc. Natl Acad. Sci. 107, 11234–11239. Web of Science CrossRef CAS PubMed Google Scholar
Jolliffe, I. T. & Cadima, J. (2016). Philos. Trans. R. Soc. A. 374, 20150202. Web of Science CrossRef Google Scholar
Kameda, H., Usugi, S., Kobayashi, M., Fukui, N., Lee, S., Hongo, K., Mizobata, T., Sekiguchi, Y., Masaki, Y., Kobayashi, A., Oroguchi, T., Nakasako, M., Takayama, Y., Yamamoto, M. & Kawata, Y. (2017). J. Biochem. 161, 55–65. Web of Science CrossRef CAS PubMed Google Scholar
Kameshima, T., Ono, S., Kudo, T., Ozaki, K., Kirihara, Y., Kobayashi, K., Inubushi, Y., Yabashi, M., Horigome, T., Holland, A., Holland, K., Burt, D., Murao, H. & Hatsui, T. (2014). Rev. Sci. Instrum. 85, 033110. Web of Science CrossRef PubMed Google Scholar
Kimura, T., Joti, Y., Shibuya, A., Song, C., Kim, S., Tono, K., Yabashi, M., Tamakoshi, M., Moriya, T., Oshima, T., Ishikawa, T., Bessho, Y. & Nishino, Y. (2014). Nat. Commun. 5, 3052. Web of Science CrossRef PubMed Google Scholar
Kobayashi, A., Sekiguchi, Y., Oroguchi, T., Okajima, K., Fukuda, A., Oide, M., Yamamoto, M. & Nakasako, M. (2016b). J. Synchrotron Rad. 23, 975–989. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kobayashi, A., Sekiguchi, Y., Oroguchi, T., Yamamoto, M. & Nakasako, M. (2018a). Sci. Rep. 8, 831. Web of Science CrossRef PubMed Google Scholar
Kobayashi, A., Sekiguchi, Y., Takayama, Y., Oroguchi, T. & Nakasako, M. (2014). Opt. Express, 22, 27892–27909. Web of Science CrossRef PubMed Google Scholar
Kobayashi, A., Sekiguchi, Y., Takayama, Y., Oroguchi, T., Shirahama, K., Torizuka, Y., Manoda, M., Nakasako, M. & Yamamoto, M. (2016a). Rev. Sci. Instrum. 87, 053109. Web of Science CrossRef PubMed Google Scholar
Kobayashi, A., Takayama, Y., Hirakawa, T., Okajima, K., Oide, M., Oroguchi, T., Inui, Y., Yamamoto, M., Matsunaga, S. & Nakasako, M. (2021). Sci. Rep. 11, 3877. Web of Science CrossRef PubMed Google Scholar
Kobayashi, A., Takayama, Y., Okajima, K., Oide, M., Yamamoto, T., Sekiguchi, Y., Oroguchi, T., Nakasako, M., Kohmura, Y., Yamamoto, M., Hoshi, T. & Torizuka, Y. (2018b). J. Synchrotron Rad. 25, 1803–1818. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kodama, W. & Nakasako, M. (2011). Phys. Rev. E, 84, 021902. Web of Science CrossRef Google Scholar
Kraft, P., Bergamaschi, A., Broennimann, Ch., Dinapoli, R., Eikenberry, E. F., Henrich, B., Johnson, I., Mozzanica, A., Schlepütz, C. M., Willmott, P. R. & Schmitt, B. (2009). J. Synchrotron Rad. 16, 368–375. Web of Science CrossRef CAS IUCr Journals Google Scholar
Kuglin, C. D. & Hines, D. C. (1975). Proceedings of the IEEE International Conference on Cybernetics and Society, New York, NY, USA, pp. 163–165. Google Scholar
Loh, N. D., Hampton, C. Y., Martin, A. V., Starodub, D., Sierra, R. G., Barty, A., Aquila, A., Schulz, J., Lomb, L., Steinbrener, J., Shoeman, R. L., Kassemeyer, S., Bostedt, C., Bozek, J., Epp, S. W., Erk, B., Hartmann, R., Rolles, D., Rudenko, A., Rudek, B., Foucar, L., Kimmel, N., Weidenspointner, G., Hauser, G., Holl, P., Pedersoli, E., Liang, M., Hunter, M. M., Gumprecht, L., Coppola, N., Wunderer, C., Graafsma, H., Maia, F. R. N. C., Ekeberg, T., Hantke, M., Fleckenstein, H., Hirsemann, H., Nass, K., White, T. A., Tobias, H. J., Farquar, G. R., Benner, W. H., HauRiege, S. P., Reich, C., Hartmann, A., Soltau, H., Marchesini, S., Bajt, S., Barthelmess, M., Bucksbaum, P., Hodgson, K. O., Strüder, L., Ullrich, J., Frank, M., Schlichting, I., Chapman, H. N. & Bogan, M. J. (2012). Nature, 486, 513–517. Web of Science CrossRef CAS PubMed Google Scholar
Lunin, V. Yu. & Woolfson, M. M. (1993). Acta Cryst. D49, 530–533. CrossRef CAS Web of Science IUCr Journals Google Scholar
MacQueen, J. (1967). Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, University of California, USA, Vol. 1, pp. 281–297. University of California Press. Google Scholar
Marchesini, S., He, H., Chapman, H. N., HauRiege, S. P., Noy, A., Howells, M. R., Weierstall, U. & Spence, J. C. H. (2003). Phys. Rev. B, 68, 140101. Web of Science CrossRef Google Scholar
Martin, A. V., Loh, N. D., Hampton, C. Y., Sierra, R. G., Wang, F., Aquila, A., Bajt, S., Barthelmess, M., Bostedt, C., Bozek, J. D., Coppola, N., Epp, S. W., Erk, B., Fleckenstein, H., Foucar, L., Frank, M., Graafsma, H., Gumprecht, L., Hartmann, A., Hartmann, R., Hauser, G., Hirsemann, H., Holl, P., Kassemeyer, S., Kimmel, N., Liang, M., Lomb, L., Maia, F. R. N. C., Marchesini, S., Nass, K., Pedersoli, E., Reich, C., Rolles, D., Rudek, B., Rudenko, A., Schulz, J., Shoeman, R. L., Soltau, H., Starodub, D., Steinbrener, J., Stellato, F., Strüder, L., Ullrich, J., Weidenspointner, G., White, T. A., Wunderer, C. B., Barty, A., Schlichting, I., Bogan, M. J. & Chapman, H. N. (2012). Opt. Express, 20, 13501–13512. Web of Science CrossRef CAS PubMed Google Scholar
Miao, J., Charalambous, P., Kirz, J. & Sayre, D. (1999). Nature, 400, 342–344. Web of Science CrossRef CAS Google Scholar
Miao, J., Chen, C.C., Song, C., Nishino, Y., Kohmura, Y., Ishikawa, T., RamunnoJohnson, D., Lee, T.K. & Risbud, S. H. (2006). Phys. Rev. Lett. 97, 215503. Web of Science CrossRef PubMed Google Scholar
Miao, J., Hodgson, K. O., Ishikawa, T., Larabell, C. A., LeGros, M. A. & Nishino, Y. (2003b). Proc. Natl Acad. Sci. USA, 100, 110–112. Web of Science CrossRef PubMed CAS Google Scholar
Miao, J., Ishikawa, T., Anderson, E. H. & Hodgson, K. O. (2003a). Phys. Rev. B, 67, 174104. Web of Science CrossRef Google Scholar
Miao, J., Ishikawa, T., Robinson, I. K. & Murnane, M. M. (2015). Science, 348, 530–535. Web of Science CrossRef CAS PubMed Google Scholar
Nakasako, M. (2018). Xray Diffraction Imaging of Biological Cells, Vol. 210 of Springer Series in Optical Sciences. Tokyo: Springer Nature. Google Scholar
Nakasako, M., Kobayashi, A., Takayama, Y., Asakura, K., Oide, M., Okajima, K., Oroguchi, T. & Yamamoto, M. (2020). Biophys Rev, 12, 541–567. CrossRef PubMed Google Scholar
Nakasako, M., Takayama, Y., Oroguchi, T., Sekiguchi, Y., Kobayashi, A., Shirahama, K., Yamamoto, M., Hikima, T., Yonekura, K., MakiYonekura, S., Kohmura, Y., Inubushi, Y., Takahashi, Y., Suzuki, A., Matsunaga, S., Inui, Y., Tono, K., Kameshima, T., Joti, Y. & Hoshi, T. (2013). Rev. Sci. Instrum. 84, 093705. Web of Science CrossRef PubMed Google Scholar
Nam, D., Park, J., GallagherJones, M., Kim, S., Kim, S., Kohmura, Y., Naitow, H., Kunishima, N., Yoshida, T., Ishikawa, T. & Song, C. (2013). Phys. Rev. Lett. 110, 098103. Web of Science CrossRef PubMed Google Scholar
Nishino, Y., Takahashi, Y., Imamoto, N., Ishikawa, T. & Maeshima, K. (2009). Phys. Rev. Lett. 102, 018101. Web of Science CrossRef PubMed Google Scholar
Oide, M., Hikima, T., Oroguchi, T., Kato, T., Yamaguchi, Y., Yoshihara, S., Yamamoto, M., Nakasako, M. & Okajima, K. (2020). FEBS J. 287, 1612–1625. Web of Science CrossRef CAS PubMed Google Scholar
Oide, M. & Nakasako, M. (2021). Sci. Rep. 11, 2827. Web of Science CrossRef PubMed Google Scholar
Oide, M., Sekiguchi, Y., Fukuda, A., Okajima, K., Oroguchi, T. & Nakasako, M. (2018). J. Synchrotron Rad. 25, 1379–1388. Web of Science CrossRef CAS IUCr Journals Google Scholar
Oroguchi, T. & Nakasako, M. (2013). Phys. Rev. E87, 022712. Google Scholar
Oroguchi, T., Sekiguchi, Y., Kobayashi, A., Masaki, Y., Fukuda, A., Hashimoto, S., Nakasako, M., Ichikawa, Y., Kurumizaka, H., Shimizu, M., Inui, Y., Matsunaga, S., Kato, T., Namba, K., Yamaguchi, K., Kuwata, K., Kameda, H., Fukui, N., Kawata, Y., Kameshima, T., Takayama, Y., Yonekura, K. & Yamamoto, M. (2015). J. Phys. B At. Mol. Opt. Phys. 48, 184003. Web of Science CrossRef Google Scholar
Oroguchi, T., Yoshidome, T., Yamamoto, T. & Nakasako, M. (2018). Nano Lett. 18, 5192–5197. Web of Science CrossRef CAS PubMed Google Scholar
Park, H. J., Loh, N. D., Sierra, R. G., Hampton, C. Y., Starodub, D., Martin, A. V., Barty, A., Aquila, A., Schulz, J., Steinbrener, J., Shoeman, R. L., Lomb, L., Kassemeyer, S., Bostedt, C., Bozek, J., Epp, S. W., Erk, B., Hartmann, R., Rolles, D., Rudenko, A., Rudek, B., Foucar, L., Kimmel, N., Weidenspointner, G., Hauser, G., Holl, P., Pedersoli, E., Liang, M., Hunter, M. S., Gumprecht, L., Coppola, N., Wunderer, C., Graafsma, H., Maia, F. R. N. C., Ekeberg, T., Hantke, M., Fleckenstein, H., Hirsemann, H., Nass, K., Tobias, H. J., Farquar, G. R., Benner, W. H., HauRiege, S., Reich, C., Hartmann, A., Soltau, H., Marchesini, S., Bajt, S., Barthelmess, M., Strueder, L., Ullrich, J., Bucksbaum, P., Frank, M., Schlichting, I., Chapman, H. N., Bogan, M. J. & Elser, V. (2013). Opt. Express, 21, 28729. Web of Science CrossRef PubMed Google Scholar
Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M. & Duchesnay, E. (2011). J. Mach. Learn. Res. 12, 2825–2830. Google Scholar
Perrakis, A., Sixma, T. K., Wilson, K. S. & Lamzin, V. S. (1997). Acta Cryst. D53, 448–455. CrossRef CAS Web of Science IUCr Journals Google Scholar
Pfeiffer, F. (2018). Nat. Photon. 12, 9–17. Web of Science CrossRef CAS Google Scholar
Rodriguez, J. A., Xu, R., Chen, C.C., Zou, Y. & Miao, J. (2013). J. Appl. Cryst. 46, 312–318. Web of Science CrossRef CAS IUCr Journals Google Scholar
Rosenthal, P. B. & Henderson, R. (2003). J. Mol. Biol. 333, 721–745. Web of Science CrossRef PubMed CAS Google Scholar
Schot, G. van der, Svenda, M., Maia, F. R. N. C., Hantke, M., DePonte, D. P., Seibert, M. M., Aquila, A., Schulz, J., Kirian, R., Liang, M., Stellato, F., Iwan, B., Andreasson, J., Timneanu, N., Westphal, D., Almeida, F. N., Odic, D., Hasse, D., Carlsson, G. H., Larsson, D. S. D., Barty, A., Martin, A. V., Schorb, S., Bostedt, C., Bozek, J. D., Rolles, D., Rudenko, A., Epp, S., Foucar, L., Rudek, B., Hartmann, R., Kimmel, N., Holl, P., Englert, L., Duane Loh, N., Chapman, H. N., Andersson, I., Hajdu, J. & Ekeberg, T. (2015). Nat. Commun. 6, 5704. Web of Science PubMed Google Scholar
Seibert, M. M., Ekeberg, T., Maia, F. R. N. C., Svenda, M., Andreasson, J., Jönsson, O., Odić, D., Iwan, B., Rocker, A., Westphal, D., Hantke, M., DePonte, D. P., Barty, A., Schulz, J., Gumprecht, L., Coppola, N., Aquila, A., Liang, M., White, T. A., Martin, A., Caleman, C., Stern, S., Abergel, C., Seltzer, V., Claverie, J.M., Bostedt, C., Bozek, J. D., Boutet, S., Miahnahri, A. A., Messerschmidt, M., Krzywinski, J., Williams, G., Hodgson, K. O., Bogan, M. J., Hampton, C. Y., Sierra, R. G., Starodub, D., Andersson, I., Bajt, S., Barthelmess, M., Spence, J. C. H., Fromme, P., Weierstall, U., Kirian, R., Hunter, M., Doak, R. B., Marchesini, S., HauRiege, S. P., Frank, M., Shoeman, R. L., Lomb, L., Epp, S. W., Hartmann, R., Rolles, D., Rudenko, A., Schmidt, C., Foucar, L., Kimmel, N., Holl, P., Rudek, B., Erk, B., Hömke, A., Reich, C., Pietschner, D., Weidenspointner, G., Strüder, L., Hauser, G., Gorke, H., Ullrich, J., Schlichting, I., Herrmann, S., Schaller, G., Schopper, F., Soltau, H., Kühnel, K.U., Andritschke, R., Schröter, C.D., Krasniqi, F., Bott, M., Schorb, S., Rupp, D., Adolph, M., Gorkhover, T., Hirsemann, H., Potdevin, G., Graafsma, H., Nilsson, B., Chapman, H. N. & Hajdu, J. (2011). Nature, 470, 78–81. Web of Science CrossRef CAS PubMed Google Scholar
Sekiguchi, Y., Hashimoto, S., Kobayashi, A., Oroguchi, T. & Nakasako, M. (2017). J. Synchrotron Rad. 24, 1024–1038. Web of Science CrossRef CAS IUCr Journals Google Scholar
Sekiguchi, Y., Oroguchi, T. & Nakasako, M. (2016). J. Synchrotron Rad. 23, 312–323. Web of Science CrossRef IUCr Journals Google Scholar
Sekiguchi, Y., Oroguchi, T., Takayama, Y. & Nakasako, M. (2014a). J. Synchrotron Rad. 21, 600–612. Web of Science CrossRef CAS IUCr Journals Google Scholar
Sekiguchi, Y., Yamamoto, M., Oroguchi, T., Takayama, Y., Suzuki, S. & Nakasako, M. (2014b). J. Synchrotron Rad. 21, 1378–1383. Web of Science CrossRef IUCr Journals Google Scholar
Shapiro, D., Thibault, P., Beetz, T., Elser, V., Howells, M., Jacobsen, C., Kirz, J., Lima, E., Miao, H., Neiman, A. M. & Sayre, D. (2005). Proc. Natl Acad. Sci. USA, 102, 15343–15346. Web of Science CrossRef PubMed CAS Google Scholar
Stuhrmann, H. B. & Miller, A. (1978). J. Appl. Cryst. 11, 325–345. CrossRef CAS IUCr Journals Web of Science Google Scholar
Svergun, D. I., Petoukhov, M. V. & Koch, M. H. J. (2001). Biophys. J. 80, 2946–2953. Web of Science CrossRef PubMed CAS Google Scholar
Takahashi, Y., Suzuki, A., Zettsu, N., Oroguchi, T., Takayama, Y., Sekiguchi, Y., Kobayashi, A., Yamamoto, M. & Nakasako, M. (2013). Nano Lett. 13, 6028–6032. Web of Science CrossRef CAS PubMed Google Scholar
Takayama, Y., Tsusaka, Y., Kagoshima, Y., Kuwamoto, S., Urushihara, Y., Li, L., Nose, S., Sudo, T., Yoshimura, M., Yokoyama, K. & Matsui, J. (2020). SPring8/SACLA Annual Report FY2018, pp. 132–135. SPring8, Sayogun, Hyogo, Japan. Google Scholar
Takayama, Y., Fukuda, K., Kawashima, M., Aoi, Y., Shigematsu, D., Akada, T., Ikeda, T. & Kagoshima, Y. (2021). Commun. Phys. 4, 48. Web of Science CrossRef Google Scholar
Takayama, Y., Inui, Y., Sekiguchi, Y., Kobayashi, A., Oroguchi, T., Yamamoto, M., Matsunaga, S. & Nakasako, M. (2015a). Plant Cell Physiol. 56, 1272–1286. Web of Science CrossRef CAS PubMed Google Scholar
Takayama, Y., MakiYonekura, S., Oroguchi, T., Nakasako, M. & Yonekura, K. (2015b). Sci. Rep. 5, 8074. Web of Science CrossRef PubMed Google Scholar
Takayama, Y. & Nakasako, M. (2012). Rev. Sci. Instrum. 83, 054301. Web of Science CrossRef PubMed Google Scholar
Takayama, Y., Takami, Y., Fukuda, K., Miyagawa, T. & Kagoshima, Y. (2018). J. Synchrotron Rad. 25, 1229–1237. Web of Science CrossRef CAS IUCr Journals Google Scholar
Takayama, Y. & Yonekura, K. (2016). Acta Cryst. A72, 179–189. Web of Science CrossRef IUCr Journals Google Scholar
Tono, K., Togashi, T., Inubushi, Y., Sato, T., Katayama, T., Ogawa, K., Ohashi, H., Kimura, H., Takahashi, S., Takeshita, K., Tomizawa, H., Goto, S., Ishikawa, T. & Yabashi, M. (2013). New J. Phys. 15, 083035. Web of Science CrossRef Google Scholar
Toyoshima, C., Nakasako, M., Nomura, H. & Ogawa, H. (2000). Nature, 405, 647–655. Web of Science CrossRef PubMed CAS Google Scholar
Uezu, S., Yamamoto, T., Oide, M., Takayama, Y., Okajima, K., Kobayashi, A., Yamamoto, M. & Nakasako, M. (2023). Sci. Rep. 13, 10802. Web of Science CrossRef PubMed Google Scholar
Wang, X., Yang, D.P., Huang, P., Li, M., Li, C., Chen, D. & Cui, D. (2012). Nanoscale, 4, 7766–7772. Web of Science CrossRef CAS PubMed Google Scholar
Williams, G. J., Pfeifer, M. A., Vartanyants, I. A. & Robinson, I. K. (2003). Phys. Rev. Lett. 90, 175501. Web of Science CrossRef PubMed Google Scholar
Xu, R., Jiang, H., Song, C., Rodriguez, J. A., Huang, Z., Chen, C.C., Nam, D., Park, J., GallagherJones, M., Kim, S., Kim, S., Suzuki, A., Takayama, Y., Oroguchi, T., Takahashi, Y., Fan, J., Zou, Y., Hatsui, T., Inubushi, Y., Kameshima, T., Yonekura, K., Tono, K., Togashi, T., Sato, T., Yamamoto, M., Nakasako, M., Yabashi, M., Ishikawa, T. & Miao, J. (2014). Nat. Commun. 5, 4061. Web of Science CrossRef PubMed Google Scholar
Yoshida, S., Harada, K., Uezu, S., Takayama, Y. & Nakasako, M. (2024). J. Synchrotron Rad. 31. Accepted. [YN5104]. Google Scholar
This is an openaccess article distributed under the terms of the Creative Commons Attribution (CCBY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.