{"publication":{"abstract":"The diffuse scattering pattern in X-ray diffraction data from protein crystals encodes the correlated motions of atoms, information needed to bridge a critical gap in our understanding of how structure gives rise to function. Although diffuse scattering is ubiquitous, it has been measured and analyzed in only a handful of cases, making it difficult to predict what impact routine analysis could have at scale. One obstacle to widespread adoption at scale is reproducibility: the diffuse signal is weak relative to Bragg diffraction and sensitive to the beamline, sample handling, and data processing. Prior studies assessing agreement between measurements have been tested only within a single group's experiments at a single facility. To test reproducibility across these settings, two independent teams measured diffuse scattering from SARS-CoV-2 Nsp3 macrodomain (Mac1) crystals at different synchrotron beamlines, using different detectors and image formats that required processing optimized for each setup, and refined their experimental practices over six beamtimes. To compare datasets quantitatively, we developed a statistical test of agreement that evaluates the correlation coefficient between datasets relative to the signal-to-noise limit in each resolution shell. By this measure (CC$_{\\text{lim}}$), both teams obtained high-quality diffuse maps that were reproducible across repeated visits and agreed between teams within experimental error. Analysis of our less successful datasets revealed the experimental variables that most limit agreement, spanning sample preparation and handling, beamline settings, and data processing choices such as the number of crystals merged. These reproducible maps lay a foundation for analyzing correlated motion in Mac1 and for interpreting perturbations such as ligand binding and temperature. More broadly, by showing that separate research groups can reproduce each other's measurements, this work moves diffuse scattering closer to becoming a routine tool in structural biology.","body":"# **Iterating diffuse scattering protocols in parallel leads to reproducible, high-quality data from Mac1**\n\n# **Introduction**\n\nA protein’s conformational dynamics often underlie its biological function, and their accurate determination remains a frontier challenge in structural biology [](https://doi.org/10.1146/annurev-biodatasci-102423-011435) [](https://doi.org/10.1038/s41592-022-01760-4) [](https://doi.org/10.1021/acs.biochem.1c00420) with direct consequences for drug design and protein engineering. The rapid progress of machine learning and generative AI methods for structure prediction [](https://doi.org/10.1038/s41586-024-07487-w) raises the possibility that similar approaches could predict conformational ensembles directly. However, current approaches remain limited [](https://doi.org/10.82153/JKXJ-TW08), in large part due to the scarcity of experimental training data that directly describe dynamics [](https://doi.org/10.1038/s41592-026-03084-z). The Protein Data Bank (PDB), which provides the foundation for structure-prediction methods, contains hundreds of thousands of deposited structures [](https://doi.org/10.3390/biom12101425) and comparatively little information about the conformational ensembles from which they arise. Most structures in the PDB are derived from X-ray crystallography, which for much of its 50-year history has prioritized high-resolution determination of average structures, resulting in a largely static view of proteins shaped by crystal packing and cryogenic data collection temperatures. Yet, protein motions occur in crystals as evidenced by subtle features present in the raw data. Atomic displacement parameters (ADPs) and alternate conformations captured during structure refinement reveal signatures of local and collective motion [](https://doi.org/10.1016/j.jmb.2007.01.032) [](https://doi.org/10.1073/pnas.0701204104) [](https://doi.org/10.1038/nmeth.2592). In addition to efforts to discover and validate alternative conformations at scale [](https://doi.org/10.82153/PFF0-CK46), there is a need for alternative experimental data sources that report directly on ensemble features.\n\nX-ray diffraction images contain two signals that encode complementary structural information: sharp Bragg peaks represent the average electron density sampled by proteins in the lattice, and the continuous diffuse scattering signal represents deviations from the average. More specifically, diffuse scattering arises from the correlated deviations of atoms from their average positions and reports on such motions across multiple length scales, from the crystal lattice to individual protein molecules [](https://doi.org/10.1038/s41467-020-14933-6) [](https://doi.org/10.1107/S2052252518001124) [](https://doi.org/10.1107/S1399004715013838). These features raise the intriguing possibility that analyzing the raw crystallography data to extract protein motions could address this training-data shortfall, advancing not only AI/ML prediction but the broader goal of understanding protein function. While diffuse scattering has long been known to encode correlated motions directly [](https://doi.org/10.1021/acs.chemrev.6b00790), a major barrier to broader adoption of diffuse scattering, however, has been the difficulty of measuring the signal with the accuracy and reproducibility routinely achieved for Bragg diffraction.\n\nBecause diffuse scattering measurements are highly sensitive to experimental conditions and data processing choices, establishing reproducibility and generalizability across beamlines has been a longstanding priority. An early study of staphylococcal nuclease marked one of the first attempts to place diffuse scattering measurements on quantitative footing, mapping the diffuse signal in three dimensions and testing its reproducibility between independent crystals [](https://doi.org/10.1073/pnas.94.12.6180). After subtraction of the isotropic component of the diffuse signal and symmetry averaging, maps from two independent crystals agreed with an overall correlation coefficient (CC) of 0.93 over a resolution range of 10 - 2.5 Å, showing that the observed features were not crystal-specific artifacts [](https://doi.org/10.1073/pnas.94.12.6180). Subsequent molecular dynamics simulations recapitulated the general features of these maps [](https://doi.org/10.1073/pnas.1416744111). However, a CC computed over a broad resolution range is dominated by the strongest features and cannot reveal where agreement breaks down. Later studies evaluated agreement as a function of resolution and demonstrated that internal consistency within a single dataset is insufficient to establish measurement reproducibility [](https://doi.org/10.1063/4.0000087). More recently, scaling algorithms that exploit high data redundancy have improved reproducibility by correcting experimental artifacts in individual measurements. Applied to data from two large insulin crystals, these methods reduced differences between independent measurements to the level of internal consistency [](https://doi.org/10.1107/S2059798324002705).\n\nAchieving high-quality diffuse scattering measurements remains experimentally challenging because the diffuse signal is weak relative to intense Bragg peaks and background scattering. Near Bragg positions, diffuse scattering forms halos that report on long-range lattice dynamics, and resolving these features requires separating them from Bragg reflections. This is facilitated by modern detectors with low noise, high dynamic range, and sharp point-spread functions [](https://doi.org/10.1016/j.sbi.2018.01.009), as well as low-mosaicity crystals that minimize Bragg peak broadening [](https://doi.org/10.1038/s41467-020-14933-6). Between Bragg peaks, weaker diffuse features contain information about protein motions but must be distinguished from background scattering arising from the beamline, air, and sample mount. Recent advances in experimental protocols and processing software have addressed these challenges, with studies of lysozyme achieving Bragg-like internal consistency (CC$_{1/2}$ ≈ 1) at fine reciprocal-space sampling and enabling quantitative modeling of both sharp halos and broad cloudy features [](https://doi.org/10.1038/s41467-020-14933-6) [](https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380). Despite these advances, the quality and reproducibility of diffuse scattering measurements have not been systematically established across different proteins, experimental teams, and beamlines. Understanding which experimental variables limit reproducibility, and how much agreement is needed to establish confidence in the resulting data, remains critical for developing diffuse scattering into a broadly applicable method for studying protein dynamics.\n\nHere, we tested the reproducibility and generalizability of protein diffuse scattering measurements as part of an intensive, multi-institution collaboration ([The Diffuse Project](https://prismscience.org/science/accelerating-dynamic-data-collection#diffuse-scattering)). As an initial target, we selected the SARS-CoV-2 Nsp3 macrodomain Mac1, which has biomedically relevant conformational dynamics [](https://doi.org/10.1126/sciadv.abf8711) and forms large crystals with high diffraction quality [](https://doi.org/10.1126/sciadv.abo5083). Two independent teams at Cornell (East) and UC San Francisco (West) collected diffuse scattering for more than 40 Mac1 crystals across six synchrotron visits at the Cornell High Energy Synchrotron Source (CHESS) beamline ID7B2 and Advanced Light Source (ALS) beamline 8.3.1. Detailed experimental conditions and data for every sample were recorded and [shared publicly](https://diffuse.science/logbook/beamtime), regardless of outcome. Using this collection of data, we evaluated reproducibility within and between teams and the effect of experimental parameters on data quality. We describe four high-quality three-dimensional diffuse maps of Mac1 that were reproducible across each team's visits and agreed between teams within the limits of experimental error. To establish a quantitative metric of agreement, we introduce CC$_{\\text{lim}}$. CC$_{\\text{lim}}$ is the maximum (limiting) correlation achievable from two independent measurements, given the signal-to-noise of each dataset. A retrospective analysis of all datasets collected, including those of lesser quality, identified the experimental variables that most strongly impact data quality and agreement for Mac1: the number of crystals merged and their orientation, humidity control during sample mounting, X-ray dose, and overall crystal size. Together, these results establish the procedures and statistical methods that will inform future studies of biologically relevant perturbations in Mac1 and other systems.\n\n# **Results**\n\n## **Optimized data collection procedure**\n\nThe East (Cornell) and West (UCSF) teams worked in parallel over multiple synchrotron visits to refine the data collection protocol. To maintain close control over beamline configuration and sample handling, we selected beamlines near each team's home laboratory. The East team worked at CHESS beamline ID7B2 with an Eiger2 X 16M detector, while the West team worked at ALS 8.3.1 with a Pilatus3 S 6M detector. Importantly, both beamlines had large, direct photon-counting detectors, which are well-suited to diffuse scattering. We collected data in shutterless mode enabling rapid continuous-rotation data collection with fine-sliced frames. The experimental setup at each beamline was adapted for low-dose room temperature data collection from large crystals, following strategies developed previously at CHESS [](https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380). With modest adjustment, both beamlines were able to provide a ~100 µm diameter X-ray beam of 14 keV with comparably low divergence, narrow energy bandwidth, and a flux of approximately 10$^{11}$ photons per second. Because diffuse scattering competes with the background from air and other materials in the beam path, extra care was taken to eliminate any stray scattering from the beamstop or upstream apertures.\n\nThe Mac1 protein construct was expressed and purified at UCSF and microcrystal seed stocks were prepared as described previously [](https://doi.org/10.1126/sciadv.abf8711). Frozen aliquots of protein, seed stocks, and crystallization reagents were shipped to Cornell, and optimization trays were set up at both sites following similar protocols, adapted to match the infrastructure available at each lab. Both protocols reliably produced large single crystals with a maximum dimension of 200-500 µm suitable for low-dose diffuse data collection. Crystals were harvested at the beamline in a humidity-controlled environment to prevent dehydration, which could reduce diffraction quality and affect reproducibility. Crystals were then protected using plastic capillaries with reservoir solution in the tip. After collecting a dataset from the crystal, the sample was translated out of the beam, and a second \"background\" dataset was collected, which includes scattering from the capillary and air around the crystal. These background datasets had the same rotation range as the crystal datasets to account for angle-dependent variations in the capillary material and shadows cast by the sample mount [](https://doi.org/10.1038/s41467-020-14933-6).\n\nData collection parameters were chosen to distribute a consistent X-ray dose over approximately 720 degrees of fine-sliced rotation images (0.1 degree per frame). In the initial experiments reported here, consistent dose proved difficult to establish between sites and synchrotron visits. In early experiments at ALS for instance, dose estimates based on total flux and pinhole aperture diameter were inaccurate due to tighter than expected focusing. Because radiation damage is a possible source of correlated disorder, and hence sample-to-sample variation in diffuse scattering, we applied an additional criterion to empirically match the dose in post-processing. Using the B-factor decay in Bragg scaling as a proxy for damage, the datasets were truncated until the decay was less than 2 Å$^{2}$ over the entire dataset. As a result, the datasets we report here are based on processing fewer than 7200 frames, but result from a consistent overall dose ([Table 1](#table1)).\n\n::::::figure{#table1 type=\"table\" label=\"Table 1\"}\n|           |                |              | **Frames** | **Mosaicity (degrees)** | **Resolution (Å)** | **Unit cell axes** |           | **Outer shell statistics (1.27-1.25 Å)** |                  |\n| --------- | -------------- | ------------ | ---------- | ----------------------- | ------------------ | ------------------ | --------- | ---------------------------------------- | ---------------- |\n|           |                |              |            |                         |                    | **a (Å)**          | **c (Å)** | **CC**$_{1/2}$                          | **Completeness** |\n| **ALS**   | **2025-10-15** | **Mac1_10** | 1-2800     | 0.014                   | 1.08               | 89.27              | 40.30     | 0.871                                    | 99.98%           |\n|           |                | **Mac1_12** | 1-2800     | 0.014                   | 1.05               | 89.24              | 40.22     | 0.942                                    | 100.00%          |\n|           |                | **Mac1_13** | 1-2800     | 0.014                   | 1.06               | 89.24              | 40.20     | 0.929                                    | 100.00%          |\n|           |                | **merged**   |            |                         |                    |                    |           | 0.980                                    | 100.00%          |\n|           | **2025-11-25** | **Mac1_2**  | 1-3600     | 0.014                   | 1.12               | 89.22              | 40.16     | 0.863                                    | 93.99%           |\n|           |                | **Mac1_4**  | 1-3600     | 0.009                   | 1.11               | 89.23              | 40.20     | 0.891                                    | 99.08%           |\n|           |                | **merged**   |            |                         |                    |                    |           | 0.939                                    | 100.00%          |\n| **CHESS** | **2025-09-24** | **Mac1_2**  | 1-3600     | 0.004                   | 1.05               | 88.41              | 39.91     | 0.921                                    | 100.00%          |\n|           |                | **Mac1_3**  | 1-3600     | 0.004                   | 1.02               | 88.42              | 39.94     | 0.940                                    | 100.00%          |\n|           |                | **Mac1_5**  | 1-3600     | 0.003                   | 1.02               | 88.41              | 39.89     | 0.944                                    | 100.00%          |\n|           |                | **merged**   |            |                         |                    |                    |           | 0.981                                    | 100.00%          |\n|           | **2025-11-05** | **Mac1_4**  | 1-7200     | 0.009                   | 1.02               | 88.42              | 39.95     | 0.988                                    | 100.00%          |\n|           |                | **Mac1_5**  | 1-7200     | 0.008                   | 1.02               | 88.44              | 39.97     | 0.984                                    | 100.00%          |\n|           |                | **merged**   |            |                         |                    |                    |           | 0.994                                    | 100.00%          |\n\n:::::figcaption\n**Table 1.** **Bragg data quality statistics for multi-crystal Mac1 datasets before and after merging.**\n:::::\n\n::::::\n\nAn additional challenge was to consistently obtain complete and uniform distributions of observations in reciprocal space. A single P4$_{3}$ Mac1 crystal is theoretically sufficient for complete coverage: the diffraction symmetry (Laue class) for this space group is _4/m_, consisting of a four-fold axis and inversion center, and the asymmetric unit occupies ⅛ of reciprocal space. However, the crystal morphology is such that it prefers to sit on the loop with the four-fold axis perpendicular to the rotation axis, resulting in an uneven distribution of multiplicities in reciprocal space, and in some cases a missing cone. We found this could be reliably overcome by merging data from multiple crystals looped in different orientations. Thus, datasets were collected from 2-3 nominally identical crystals during each synchrotron visit for subsequent merging.\n\n## **Four high-quality datasets establish a baseline for reproducibility**\n\nAfter converging on a shared optimized protocol for Mac1, both teams performed two independent data collections (different samples, different synchrotron visits) at their respective home synchrotron beamlines, resulting in four multi-crystal datasets. These datasets enable comparisons between independent collections at the same site and across sites. The diffraction resolution, assessed by the outer shell CC$_{1/2}$ > 0.3 criterion, varied slightly between crystals but regularly exceeded 1.1 Å, on par with the best reported P4$_{3}$ structures at ambient temperature (e.g. PDB ID [7TWF](https://www.rcsb.org/structure/7TWF) [](https://doi.org/10.1126/sciadv.abo5083)). In addition, the mosaicity estimates from DIALS were consistently 0.014˚ or lower, which is near the limit of what can be reliably measured at CHESS and ALS and are remarkably low for a protein crystal [](https://doi.org/10.1016/S0076-6879(03)68015-8). Importantly, the low mosaicity is a strong indicator that crystals did not experience dehydration stresses during mounting, and it enables diffuse scattering to be measured close to intense Bragg peaks without Bragg signal leaking into the diffuse measurement. Due to detector distance limits at CHESS, and the need to have sufficient number of pixels between Bragg peaks for separating the diffuse scattering, the detector was placed at a distance that was not optimal for measuring the highest resolution shell with 100% completeness. Therefore, the data were truncated at a consistent cutoff of 1.25 Å across all measurements ([Table 1](#table1)). For individual crystal datasets, the completeness of the outer shell was limited by the preferential orientation of crystals on the loop. However, after merging, all four datasets have an outer shell completeness of 100%, showing the benefit of a multi-crystal approach. Interestingly, the lattice parameters are not identical for crystals collected at CHESS and ALS, with ALS crystals having ~2.5% larger unit cell volume ([Table 1](#table1)). This change in dimensions likely results from the slightly different crystallization procedures applied by each team, such as differences in drop size and vapor diffusion conditions (sitting drop vs hanging drop). Nonetheless, when comparing data from crystals prepared by the same team, the lattice parameters are very consistent, suggesting that the humidity environment of the crystals was well controlled during harvesting and data collection. To assess whether crystals were truly isomorphous, we compared the outer shell CC$_{1/2}$ values for individual crystals and that of the merged dataset. We expect CC$_{1/2}$ to improve (increase) after merging if discrepancies are caused only by signal-to-noise and the crystals are truly isomorphous. The outer shell CC$_{1/2}$ increased after merging in all cases ([Table 1](#table1)), consistent with highly isomorphous crystals.\n\n## **Separating isotropic and anisotropic contributions to diffuse scattering**\n\nFollowing Bragg data processing, three-dimensional maps of diffuse scattering were computed using mdx2 [](https://doi.org/10.1107/S2059798324002705) using the same diffraction images analyzed with DIALS. Overall, the four diffuse maps are visually similar and share important features ([Figure 1](#figure1)). While there is a long history of comparing Bragg data reduction from different detector formats and beamlines, the similarity of the diffuse maps indicates the robustness of mdx2 and the underlying libraries to the differences in experimental setups. All have very little scattering at low resolution, with most of the intensity concentrated in the so-called \"water ring\", a ~3 Å resolution peak commonly observed in diffraction images of macromolecules, which in reality has contributions from both protein and solvent disorder [](https://doi.org/10.1002/prot.21246). The isotropic intensity then decays again at high resolution. On top of this isotropic scattering, two types of diffuse scattering are seen: halos and cloudy patterns. Halos are intense features observed at the Bragg peak locations that are the result of long-ranged displacement correlations between unit cells. In mdx2, voxels centered at a Bragg peak location do not include Bragg diffraction: the peak is masked out, and only photons surrounding the peak are integrated. In reciprocal space slices that pass through a plane of Bragg peaks (one with an integer Miller index _h_, _k_, or _l_), the diffuse scattering contains a lattice of diffuse halos ([Figure 1](#figure1), middle column). In reciprocal-space slices midway between the reciprocal lattice planes (where _h_, _k_, or _l_ is a half-integer), the reciprocal lattice is less obvious in the pattern ([Figure 1](#figure1), right column), suggesting that it derives predominantly from correlated motions within the unit cell, including the internal protein and solvent motion. Although these two types of scattering, halos and cloudy patterns, can be distinguished visually, in reality they overlap significantly in reciprocal space, and further modeling is required to analyze them separately [](https://doi.org/10.1038/s41467-023-36734-3). In the analysis that follows, we compare the total anisotropic scattering, excluding the points with integer Miller indices, which reduces (but does not eliminate) the influence of the intense halo scattering which would otherwise dominate the statistics.\n\n::::::figure{#figure1 align=\"center\" type=\"image\" label=\"Figure 1\"}\n\n:::::image{width=\"100%\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_b0626931_ac57d438b6f8\"}\n:::::\n\n:::::figcaption\n**Figure 1.** **Three-dimensional maps of diffuse scattering from Mac1 replicated at CHESS and ALS.** Maps of total diffuse intensity were reconstructed on a grid oversampled 2x2x4 relative to the reciprocal lattice (voxel dimensions: a*/2, b*/2, c*/4). Intensity was normalized to 1 at the maximum value of the isotropic scattering to better compare the four datasets (one per row). The datasets appear nearly identical and share common features. **Left column**, total intensity is shown a spherical surface drawn at 2 Å resolution and central sections (_h_=0, _k_=0, and _l_=0 planes). The total intensity appears as fluctuations on top of a mostly isotropic scattering with a bright ring at ~3.3 Å resolution. **Middle column**, anisotropic intensity (total minus isotropic average) in central sections showing a lattice of haloes that extend to the limit of the dataset at 1.25 Å resolution. **Right column**, the anisotropic intensity in sections passing through half-integer planes (_h_=½, _k_=½, and _l_=½) shows a mostly cloudy pattern extending to high resolution.\n:::::\n\n::::::\n\n::::::figure{#movie1 align=\"center\" type=\"iframe\" unnumbered=\"true\" label}\n\n:::::iframe{width=\"100%\" height=\"650px\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_73562a7f_c753017f8ced\"}\n:::::\n\n:::::figcaption\nAnimation: [chess_20250924_intensity_sub.mp4](https://drive.google.com/open?id=1UAWAJseBKkay9BEEuNU5kGff3FgwL2NT)\n:::::\n\n::::::\n\nThe diffuse scattering signal contains a spherically isotropic component, whose most prominent feature is the broad ~3 Å \"water ring\" peak described above. Although the isotropic component lacks information on the collective protein motions that gives rise to anisotropic scattering and is often overlooked in models focused primarily on protein motions, it contains information about the short-ranged correlations of protein and solvent atoms [](https://doi.org/10.1002/prot.21246). These signals are important for benchmarking models that include atomistic description of solvent, such as MD simulations [](https://doi.org/10.1038/s41467-020-14933-6) [](https://doi.org/10.1073/pnas.1416744111). Moreover, accurately measuring the isotropic scattering requires a well-measured background and other corrections which we aim to assess here. Overall, the four datasets show similar resolution-dependent isotropic scattering profiles ([Figure 2](#figure2)a, top panel), with a dominant peak at ~3.3 Å, a broad shoulder at ~2.2 Å, and relatively flat intensity beyond 1.7 Å (dashed lines in [Figure 2](#figure2)a, top panel). The isotropic component here includes incoherent Compton scattering, particularly at high angles, which would need to be corrected before comparing with MD simulations as described previously [](https://doi.org/10.1038/s41467-020-14933-6).\n\n::::::figure{#figure2 align=\"center\" type=\"image\" label=\"Figure 2\"}\n\n:::::image{width=\"100%\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_5418fbcd_6072efb90f0e\"}\n:::::\n\n:::::figcaption\n**Figure 2.** **Evaluating consistency between three-dimensional diffuse maps.** Intensity statistics were computed for all four high-quality datasets within narrow resolution shells and plotted as a function of scattering vector magnitude (_s_ = 1/_d_ where _d_ is the resolution). **a**, The isotropic component (top panel), and the root mean squared (R.M.S.) deviation of intensity (bottom panel) shows consistent resolution-dependent features between datasets. All intensities are normalized by the peak of the isotropic component. Vertical dashes lines mark particular resolution shells that are discussed in the main text (same in all panels). **b**, The internal data quality for each dataset is estimated using the CC* metric (left column, purple lines), computed from the correlation of random half-datasets (see main text). Similarly, we computed CC$_{\\text{pair}}$, the resolution-dependent Pearson correlation coefficient between each pair of datasets (blue lines, panels on the right comparing datasets row and column-wise in the grid). The theoretical limit for the best possible agreement is CC$_{\\text{lim}}$, the product of CC* for each dataset being compared (cyan lines). The region between CC$_{\\text{pair}}$ and CC$_{\\text{lim}}$ is shaded for emphasis. Datasets agree within the limits of signal-to-noise, except at resolutions below ~6 Å where the signal is relatively weak (bottom panel of part a).\n:::::\n\n::::::\n\nIn contrast, the anisotropic component contains both the sharp halo and broader cloudy features described above. Although anisotropic scattering varies with direction in reciprocal space, its overall magnitude can also be examined as a function of resolution [](https://doi.org/10.1107/S1399004715013838). Here, we estimate this magnitude from the mean-squared deviation of intensity within each resolution shell after subtracting the expected contribution from photon shot-noise, which becomes significant at high resolution. The resolution-dependent profile and overall magnitude are consistent between the four datasets ([Figure 2](#figure2)a, bottom panel). Similar to the mean isotropic intensity, the anisotropic signal is small at low resolution, rises to a maximum in the water ring, features a shoulder at moderate resolution, and gradually decays at high resolution. Although the resolution-dependence is qualitatively similar for anisotropic and isotropic signals, the peak locations do not coincide exactly, especially the main peak in the anisotropic signal, which is shifted toward lower resolution than the corresponding peak of the isotropic component ([Figure 2](#figure2)a, dashed line at 3.3 Å). This difference indicates that the two signals likely have different physical origins.\n\n## **Assessing internal data quality of individual datasets**\n\nBecause diffuse scattering is a weak signal that can be obscured by noise and measurement errors, before datasets can be compared with each other, we first need a measure of internal data quality, or consistency. For diffuse scattering, the signal-to-noise level tends to vary strongly with resolution, both because the signal strength varies (i.e. the anisotropic signal in [Figure 2](#figure2)a, bottom panel) and because the number of photons per voxel falls off rapidly with resolution due to the data collection geometry. Consequently, there is no single number that can capture the complete picture of data quality. First, we compute CC$_{1/2}$ within each resolution shell as a direct measure of internal consistency between random half-datasets. From CC$_{1/2}$, we obtain another measure of data quality, CC*, which represents the maximum expected correlation between the measured data and a perfect model, given the noise in the data [](https://doi.org/10.1126/science.1218231) [](https://doi.org/10.1038/s41467-020-14933-6). In this way, CC* translates measurement precision into an upper bound on the model-data agreement that can realistically be achieved. For all four datasets, CC* ([Figure 2](#figure2)b, purple curves) tracks the resolution dependence of the anisotropic signal strength ([Figure 2](#figure2)a, bottom panel): rising at low resolution to exceed 0.85 between 6 and 2.2 Å and approaching 1 near 4 Å (CC* = 0.965–0.997), then decreasing at resolutions beyond 1.7 Å. Due mainly to differences in total illuminated volume (i.e. crystal size and number of crystals), the four datasets have different overall signal-to-noise levels, resulting in somewhat different CC* values. The CHESS 2025-11-05 dataset has the highest signal-to-noise, consistent with the exceptionally large crystals used for data collection (two crystals approximately 500 µm in diameter). Notably, this dataset maintains CC* above 0.95 across a broad resolution range of 13.3–1.8 Å.\n\n## **Comparing data quality and consistency across datasets**\n\nTo compare datasets to each other, we next computed the resolution-dependent correlation coefficients between anisotropic maps. Because the observed correlation between two maps is limited both by measurement noise and by any genuine differences in the underlying diffuse signal, the pairwise map-map CC (CC$_{\\text{pair}}$) alone cannot distinguish between these effects. We therefore define a limiting map-map CC as CC$_{\\text{lim}}$, the maximum correlation expected between two datasets given the noise in each measurement, which is simply the product of their respective CC* values (see Methods for derivation). If a gap is seen between CC$_{\\text{lim}}$ and CC$_{\\text{pair}}$, this can be interpreted as a statistically significant difference in the anisotropic diffuse pattern within that resolution shell. We find that CC$_{\\text{pair}}$ closely approaches CC$_{\\text{lim}}$ for all dataset comparisons, indicating that most differences between maps can be explained by measurement noise ([Figure 2](#figure2)b, blue curves). The largest gaps between CC$_{\\text{pair}}$ and CC$_{\\text{lim}}$ appear at low resolution, below 6 Å, where the anisotropic diffuse signal is relatively weak ([Figure 2](#figure2)a, bottom panel). These differences are seen between datasets collected by the same teams (i.e. comparing ALS vs. ALS, or CHESS vs. CHESS) so they cannot be attributed to differences in sample preparation or beamline configuration between sites. These low-resolution differences notwithstanding, agreement between replicates is excellent, with no significant differences seen between East and West teams.\n\nThe four datasets reported here are comparable in quality to the foundational lysozyme datasets published previously [](https://doi.org/10.1038/s41467-020-14933-6) [](https://doi.org/10.1038/s41467-023-36734-3) and contain similar anisotropic features (halos, cloudy patterns), giving confidence that the GOODVIBES/DISCOBALL workflow [](https://doi.org/10.1038/s41467-023-36734-3) developed to model these features in lysozyme can be applied to Mac1. Moreover, the datasets establish a generalizability baseline for diffuse scattering that will enable the next experimental steps for Mac1: applying perturbations in the form of ligand binding or temperature variation, and observing the effect on collective protein motion using X-ray diffraction.\n\n## **Experimental factors leading to high data quality and reproducibility**\n\nThe strong agreement that we see between ALS and CHESS datasets, and between repeated experiments by the same team, emerged from an iterative process of refining the sample handling and data collection procedures across multiple synchrotron trips. To identify which aspects of the procedure were most important, we retrospectively analyzed less ideal datasets collected during this optimization process. These datasets allowed us to assess the importance of multi-crystal averaging, controlling humidity, limiting radiation damage, and growing large high-quality crystals. For multi-crystal averaging, we divided the CHESS 2025-09-24 dataset into subsets of one, two, or three crystals. For humidity, we used data from ALS 2025-07-01, when we varied the mounting procedure (free air or a sleeve, with and without oil) to see the impact on data quality. For radiation damage, we used data from ALS 2025-10-15, when we were working to match the CHESS dose rate. For crystal size, we used data from smaller crystals collected at ALS 2026-04-23. Collectively, these experiments map out many of the factors influencing diffuse scattering data quality.\n\n### **Multi-Crystal Datasets**\n\nComplete three-dimensional reciprocal space maps of diffuse scattering have been reported from both single- and multi-crystal data collections [](https://doi.org/10.1038/s41467-020-14933-6) [](https://doi.org/10.1038/s41467-023-36734-3) [](https://doi.org/10.1107/S2059798324002705) [](https://doi.org/10.1063/4.0000087). For a high-symmetry crystal form, a single crystal can provide complete reciprocal space coverage. For example, insulin in I2$_{1}$3 has high symmetry such that a complete dataset with high multiplicity can be obtained from a single crystal [](https://doi.org/10.1107/S2059798324002705). Merging data from multiple insulin crystals, however, further improved data quality according to CC$_{1/2}$ [](https://doi.org/10.1107/S2059798324002705). In other cases, such as lysozyme in the P1 space group, a multi-crystal approach was necessary to obtain a complete map [](https://doi.org/10.1038/s41467-020-14933-6). The number of crystals required for a diffuse dataset therefore depends on both reciprocal-space coverage and signal-to-noise. Here, we evaluate these two factors for Mac1.\n\nFocusing on the CHESS 2025-09-24 three-crystal dataset ([Table 1](#table1)), we generated diffuse scattering maps from each individual crystal and from all possible crystal combinations ([Figure 3](#figure3)a). Although individual Bragg datasets displayed high completeness ([Table 1](#table1)), the single-crystal diffuse maps display a strong propeller-like pattern caused by uneven multiplicity across reciprocal space ([Figure 3](#figure3)a, top). In the two-crystal merges, this pattern is reduced but remains visible, particularly for xtal2 + xtal3, consistent with the crystals adopting similar orientations during data collection ([Figure 3](#figure3)a, middle). In the three-crystal merge, the propeller feature is no longer apparent ([Figure 3](#figure3)a, bottom). As the orientation of this pattern depends directly on the crystal orientation during data collection, this result also illustrates how orientation bias can affect diffuse-map quality even for high-symmetry crystals.\n\n::::::figure{#figure3 align=\"center\" type=\"iframe\" label=\"Figure 3\"}\n\n:::::iframe{width=\"100%\" height=\"1000px\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_0aa91147\"}\n:::::\n\n:::::figcaption\n**Figure 3.** **The number of crystals affects the uniformity of reciprocal-space sampling and the signal-to-noise ratio. a,** The _l_=0.5 slice of the isotropic-subtracted (anisotropic) map for every combination of three crystals from CHESS 2025-09-24: each single crystal (top row), each two-crystal merge (middle row), and the three-crystal merge (bottom). Each map is normalized to its own isotropic maximum. The single-crystal maps show a propeller-like pattern arising from uneven multiplicity across reciprocal space; this pattern is reduced in the two-crystal merges and absent in the three-crystal merge. **b,** Signal-to-noise as a function of resolution for each merge, calculated as the noise-corrected anisotropic root-mean-square (RMS) intensity (as in [Figure 2](#figure2)) divided by the RMS propagated error, with Bragg voxels excluded. Curves are colored by the number of merged crystals, matching the panel titles in a. Signal-to-noise increases with crystal number, and the resolution at which each curve reaches signal = noise (dashed line) extends to higher _s_ with additional crystals.\n:::::\n\n::::::\n\nAs expected, signal-to-noise also improves with the number of crystals. We quantified this as the ratio of the noise-corrected anisotropic root-mean-square (RMS) intensity to the RMS propagated error, which increases with each additional crystal ([Figure 3](#figure3)b). Correspondingly, the resolution at which the signal falls to the level of the propagated error (RMS signal/error = 1) shifts to higher resolution with increasing number of crystals, from 1.8 - 2 Å resolution (s = ~0.5 - 0.55 Å$^{-1}$) for the single crystals to 1.6 Å (s = ~0.62 Å$^{-1}$) for the three-crystal merge. These results demonstrate that merging data from multiple crystals improves both reciprocal-space coverage and diffuse signal quality and suggest that sampling diverse crystal orientations is more important than reaching a strict minimum number of crystals. Indeed, some of our high-quality diffuse datasets described above were obtained from only two crystals ([Table 1](#table1)). Because multi-crystal merging depends on reproducibility between individual crystal datasets, we next examined experimental factors that contribute to crystal-to-crystal variation.\n\n### **Humidity**\n\nPrevious work demonstrated that high-quality diffuse scattering data can be obtained by maintaining crystal hydration in a humidity-controlled chamber, with some samples benefiting from an additional oil coating before mounting in a sleeve [](https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380). These approaches are needed to ensure that the crystal remains hydrated during sample handling and data collection at room temperature, enabling reproducible data. Because not every crystallography beamline is equipped with a humidity chamber, we explored whether other handling strategies to maintain crystal hydration could reproducibly yield high-quality Mac1 datasets. During an early ALS beamtime (2025-07-01), we varied two experimental factors: whether the crystal was dipped in oil (Paratone-N or Santovac-5) and whether it was mounted inside a sleeve. As the CHESS crystals showed a distinctly smaller unit cell, with shorter a (= b) and c axes relative to the ALS crystals, we restricted these comparisons to the ALS datasets. We attribute this systematic difference to variation in crystal growth rather than to differences in handling conditions. For these experiments, all crystals were harvested under ambient conditions without humidity control. Four combinations were tested: Paratone-N without a sleeve ([Figure 4](#figure4), red bars), Paratone-N with a sleeve ([Figure 4](#figure4), orange bars), Santovac-5 with a sleeve ([Figure 4](#figure4), green bars), and no oil with a sleeve ([Figure 4](#figure4), blue bars). Bragg data quality indicators were then used to compare these four 2025-07-01 datasets with two high-quality ALS datasets harvested with humidity control during later visits (ALS 2025-10-15 and 2025-11-25; [Figure 4](#figure4), grey bars; [Table 1](#table1)). Each 2025-07-01 condition is a single-crystal dataset. For the references, Wilson B and unit cell dimensions come from each beamtime’s multi-crystal merge, and the reference values quoted below are the mean of the two merges. Mosaicity is refined per crystal during integration, so each reference mosaicity is the median across that beamtime’s individual crystals, and the quoted reference value is the median of all five crystals ([Figure 4](#figure4), dashed lines).\n\n::::::figure{#figure4 align=\"center\" type=\"iframe\" label=\"Figure 4\"}\n\n:::::iframe{width=\"100%\" height=\"1000px\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_843e074a_21b4f8d5674e\"}\n:::::\n\n:::::figcaption\n**Figure 4.** **Crystal handling during harvesting affects Bragg data quality.** Mac1 crystals from the ALS 2025-07-01 beamtime were mounted under four conditions: no sleeve + Paratone-N, sleeve + Paratone-N, sleeve + Santovac-5, and sleeve + no oil. Rows show Wilson B, mosaicity, and unit cell a (= b) and c, with two humidity-controlled reference datasets (ALS 2025-10-15 and 2025-11-25) in grey. Values for each 2025-07-01 condition come from a single crystal. For the references, Wilson B and unit cell values come from each beamtime’s multi-crystal merge; mosaicity, which is refined per crystal during integration, is the median of that beamtime’s individual crystals. Dashed lines give the mean of the two reference datasets for Wilson B and unit cell, and the median of all five reference crystals for mosaicity. The unsleeved crystal has the highest Wilson B and the most contracted c axis, consistent with dehydration during mounting. Among the sleeved crystals, the no-oil mount is closest to the references on all four metrics, with the lowest mosaicity and the least-contracted c axis.\n:::::\n\n::::::\n\nFor Mac1 mounted in Paratone-N without a sleeve ([Figure 4](#figure4), red bars), every Bragg quality indicator was worse than in the humidity-controlled reference datasets ([Figure 4](#figure4), grey bars). The Wilson B factor, which reflects the average atomic displacement from thermal motion and disorder, was 20.5 Å², compared with a reference mean of 13.1 Å². Mosaicity, which reflects long-range crystal order, was 0.120°, compared with a reference median of 0.014°. The unit cell was also contracted relative to the references, most notably along c (39.48 Å versus 40.20 Å). A smaller unit cell is not detrimental on its own, but relative to humidity-controlled crystals this contraction indicates dehydration. Combined with the elevated Wilson B and mosaicity, these results show that a Paratone-N coating without a sleeve does not protect the crystal from dehydration during mounting and data collection.\n\nMounting in a sleeve restored Wilson B factors to reference levels in all three conditions (11.7–12.6 Å² versus a reference mean of 13.1 Å²; [Figure 4](#figure4), orange, green, and blue bars), but the effect of oil coating was more variable. With Paratone-N ([Figure 4](#figure4), orange bars), the c axis was ~0.50 Å shorter than the reference, and mosaicity was 0.022°, slightly above the reference median of 0.014°. With Santovac-5 ([Figure 4](#figure4), green bars), the c axis was ~0.55 Å shorter, and mosaicity was 0.186°, more than an order of magnitude higher than the reference. The shorter c axes of the oil-coated crystals suggest partial dehydration even inside a sleeve. This may have occurred during transfer from the drop into oil or through contact with the oil itself. The elevated mosaicity with Santovac-5 further suggests that this oil introduced lattice strain. Without oil ([Figure 4](#figure4), blue bars), all four parameters were close to the reference datasets: mosaicity was 0.009°, and both unit cell axes were within ~0.1 Å of the reference. For Mac1, oil coating therefore offers little benefit and can compromise both crystal hydration and lattice order.\n\nOverall, mounting conditions strongly affected Bragg data quality and we did not attempt to extract diffuse scattering from the inferior datasets. Only the crystal mounted in a sleeve without oil matched the reference datasets. However, it was harvested under ambient conditions, which are less controllable and can dehydrate crystals to varying degrees during handling. For our highest-quality datasets, we therefore harvested crystals with humidity control. The two humidity-controlled ALS datasets ([Figure 4](#figure4), grey bars) agree closely on every parameter across separate beamtimes, indicating that this approach is reproducible. Notably, as established in previous work, the sleeve mount has an important advantage over oil mounts as it allows a well-matched background to be collected by translating along the sleeve and using an equivalent path length of the same material [](https://doi.org/10.1038/s41467-020-14933-6) [](https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380). With oil, the amount around the crystal varies between mounts and with rotation, so there is no equivalent region from which to collect a matched background, and any mismatch will make it more difficult to extract the weak diffuse signal. We therefore strictly adopted the sleeve approach in subsequent beamtimes to improve background subtraction and obtain high-quality diffuse data.\n\n### **X-ray Dose**\n\nRadiation damage can alter diffuse scattering signals. Therefore, total X-ray dose must be accounted for when comparing datasets and should be matched across facilities. The accumulated dose depends on incident flux and per-frame exposure time, so matching dose across facilities requires exploring both parameters.  Between CHESS ID7B2 and ALS 8.3.1, this required care because the beamlines differ in beam profiles (top-hat versus Gaussian, respectively) and achievable detector frame rates. While calibrating these parameters, we collected several datasets at high dose. We compared a dataset at ALS with double the exposure per frame of our highest-quality ALS collections (0.1 s vs 0.05 s per frame, Δφ = 0.1°). As all other conditions were the same, including harvesting with humidity control, this dataset isolates the effect of X-ray dose. Over the course of the high-dose sweep, diffraction quality visibly degraded. To quantify this, we compared Bragg statistics for the high-dose and low-dose datasets. The high-dose dataset had higher mosaicity (0.056° versus 0.014°) and a modestly higher Wilson B (13.9 versus 13.2 Å²). These changes are hallmarks of radiation damage.\n\nIt was less clear how such damage would manifest in the diffuse signal. To test for dose dependence, we split each rotation series at its midpoint and merged the first half and the second half separately. The halves within each dataset (0.1 s and 0.05 s exposures) share a common scale model so each half can be compared directly. The two high-dose halves were normalized by a common factor (the average of their isotropic maxima), so intensity differences reflect genuine changes. The low-dose dataset was normalized by its own isotropic maximum so only the overall shape can be compared. Although the overall shape is similar, we see modest differences in the isotropic component when comparing the 0.1 s exposure time dataset against the 0.05 s exposure time datasets ([Figure 5](#figure5)a, left).  For both halves of the high-dose data, the shoulder intensity at 2.2 Å (_s_ ≈ 0.4 Å⁻¹) falls off more steeply than for the low-dose dataset. The anisotropic component shows larger differences, both relative to the lower-dose dataset and between the two high-dose halves ([Figure 5](#figure5)a, right). In addition, the water peak shifts to slightly lower resolution for the high-dose datasets. In the high-dose second half, the low resolution (~10 Å) RMS intensity is elevated, but the curves agree beyond ~2.2 Å (s ≈ 0.45 Å⁻¹).\n\n::::::figure{#figure5-part1 align=\"center\" type=\"iframe\" unnumbered=\"true\"}\n\n:::::iframe{width=\"100%\" height=\"1000px\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_f0783a32_0f3969943b4f\"}\n:::::\n\n::::::\n\n::::::figure{#figure5 align=\"center\" type=\"iframe\" label=\"Figure 5\"}\n\n:::::iframe{width=\"100%\" height=\"520px\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_8bc6b587_d41b0f09519e\"}\n:::::\n\n:::::figcaption\n**Figure 5.** **Radiation damage leaves a signature in the diffuse signal.** Each rotation series was split at its midpoint and the two halves were merged with a common scale model. The high-dose 0.1 s dataset is a single crystal (double the standard per-frame exposure) and the lower-dose 0.05 s dataset is the ALS 2025-10-15 multi-crystal reference merge. **a**, 1D profiles of the isotropic and anisotropic (Bragg voxels excluded) components for the full 0.05 s dataset and each 0.1 s half. Both 0.1 s halves were normalized to the average of their isotropic maxima. The isotropic profiles are essentially indistinguishable, but in the anisotropic component the 0.1 s halves diverge, with the second half carrying higher RMS intensity. **b**, _l_=0.5 slices of the anisotropic map for both halves of each dataset. The two 0.05 s halves are indistinguishable, whereas the 0.1 s halves differ in the low-resolution region. The propeller-like pattern in the 0.1 s slices reflects single-crystal coverage and is distinct from the half-to-half difference. **c**, Correlation between halves for each dataset (anisotropic component, Bragg voxels excluded). Diagonal panels show CC* for each half, a measure of its internal consistency. Off-diagonal panels compare CC$_{\\text{pair}}$ with CC$_{\\text{lim}}$ for the two halves. For the 0.05 s dataset, CC$_{\\text{pair}}$ matches CC$_{\\text{lim}}$. For the 0.1 s dataset, there is a gap between CC$_{\\text{pair}}$ and CC$_{\\text{lim}}$, indicating a difference between halves beyond noise.\n:::::\n\n::::::\n\nThe _l_=0.5 reciprocal-space slices show clear structural differences between the low-dose and high-dose data ([Figure 5](#figure5)b). Both halves of the low-dose data look nearly identical ([Figure 5](#figure5)b, top). The two high-dose halves differ from each other, most visibly at low resolution (d > ~4 Å), where features are stronger in the second half, consistent with the 1D profiles ([Figure 5](#figure5)b, bottom). Relative to the low-dose data, both high-dose halves differ in shape at low resolution and near the water ring resolution. The high-dose slices show the propeller-like pattern characteristic of the lower redundancy due to the orientation of the crystal. However, this limitation does not obscure the clear differences due to dose.\n\nTo evaluate whether these differences exceed the noise, we compared CC$_{\\text{pair}}$ for the two halves with CC$_{\\text{lim}}$ ([Figure 5](#figure5)). Here, CC$_{\\text{lim}}$ is the product of the two halves' CC* values, which is the CC expected if the halves differ only by noise. For the low-dose dataset, CC$_{\\text{pair}}$ for the two halves matches the CC$_{\\text{lim}}$ across the full resolution range, indicating no difference beyond the noise ([Figure 5](#figure5)c, left). However, for the high-dose dataset, CC$_{\\text{pair}}$ for the two halves falls slightly below CC$_{\\text{lim}}$ between 10 - 5 Å resolution, indicating that the differences seen in that resolution range in the 1D profiles and the slices are from real differences beyond the noise ([Figure 5](#figure5)c right). This gap closes around 2.5 Å and beyond, which is consistent with the 1D profiles. Establishing the physical mechanism behind these dose-dependent changes in the diffuse signal will require further experiments. For this work, the relevant conclusion is that radiation damage leaves a measurable signature, providing a diagnostic for dose effects when screening for high-quality diffuse data.\n\n### **Crystal Size**\n\nDiffuse scattering is inherently low intensity, and the options for amplifying the signal are limited. As shown above, dose must be constrained to avoid radiation damage, so the signal strength cannot simply be improved with higher flux or longer exposures. A better alternative is utilizing large crystals because a larger illuminated volume boosts the diffuse signal for a given dose. Our high-quality datasets were collected from crystals in the 200–500 µm range, with the highest CC* achieved using the largest crystals measured in this study (CHESS 2025-11-05).\n\nThe crystals available at our April 2026 ALS beamtime were roughly 100 µm. They were harvested with humidity control and the dose was tuned to avoid radiation damage, so we can isolate the effect of crystal size. We compared a three-crystal merge of the small crystals against a two-crystal merge of the largest crystals (CHESS 2025-11-05) to span the extremes of our data. This comparison spans two facilities, but as shown above our CHESS and ALS datasets agree to within their combined noise ([Figure 2](#figure2)b), so the differences here are attributable primarily to crystal size.\n\nThe small-crystal data reproduce the main features of large-crystal results, and the isotropic and anisotropic profiles track each other closely. For the isotropic curve beyond the water peak, the small-crystal curve is consistently lower relative to its maximum than the large crystal curve, especially near the shoulder at 2.2–2.5 Å (s ≈ 0.40–0.45 Å⁻¹)([Figure 6](#figure6)a). For the anisotropic component, the main peak shifts to slightly lower resolution. Part of this divergence, however, may reflect the behavior of scaling and merging on noisier data rather than a change in the underlying scattering. The _l_=0.5 reciprocal space slice shows the same overall pattern in both datasets, but the small crystal map is weaker in intensity and the features are less sharply defined ([Figure 6](#figure6)b).\n\n::::::figure{#figure6 align=\"center\" type=\"iframe\" label=\"Figure 6\"}\n\n:::::iframe{width=\"100%\" height=\"1000px\" src=\"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_67edc2dc_40cfd2e43500\"}\n:::::\n\n:::::figcaption\n**Figure 6.** **Crystal size affects diffuse signal strength. Comparison of the smallest crystals measured (ALS 2026-04-23, ~100 µm, three crystals) against the largest (CHESS 2025-11-05, 200–400 µm, two crystals). a,** 1D profiles of the isotropic and anisotropic (Bragg voxels excluded) components. The isotropic profiles are similar in shape. In the anisotropic component, the small-crystal data are lower between s ≈ 0.25 and 0.55 Å⁻¹ and level off at a higher baseline above s ≈ 0.6 Å⁻¹. **b,** _l_ = 0.5 slices of the anisotropic map. Both show the same principal features, but the small-crystal map is noticeably weaker and less sharply defined. **c.** Signal-to-noise as a function of resolution, calculated as the noise-corrected anisotropic RMS intensity (as in a) divided by the RMS propagated error, with Bragg voxels excluded. The small-crystal data have lower signal-to-noise at all resolutions and reach signal = noise (dashed line) at s ≈ 0.48 Å⁻¹, compared with ≈ 0.8 Å⁻¹ for the large crystals.\n:::::\n\n::::::\n\nThe difference in signal strength is clearest in the ratio of the noise-corrected anisotropic RMS intensity to the propagated error. The small crystal signal-to-noise ratio is lower at all resolutions, by roughly four- to five-fold, even though the small crystal merge includes three crystals and the large crystal merge only two. The small crystal signal reaches the level of the noise at ~ 2.1 Å (s ≈ 0.48 Å⁻¹), compared with ~ 1.25 Å (s ≈ 0.8 Å⁻¹) for the large crystals, resulting in a lower usable resolution range. Collectively, these results indicate that increasing crystal size is the most straightforward route to increasing the ability to obtain a reproducible diffuse scattering signal.\n\n# **Discussion**\n\nOur results show that diffuse scattering can be highly reproducible when experimental variables are carefully controlled, establishing it as a routine, generalizable measurement for structural biology. For Mac1, carefully controlling humidity during sample mounting yielded highly isomorphous datasets from crystals in different orientations, producing maps of uniform quality throughout reciprocal space. Comparing datasets collected across separate synchrotron visits showed that these measurements can be made routinely, an important step toward planned studies of how the diffuse signal changes in response to perturbations such as ligand binding, mutation, or temperature. Finally, working in parallel at two separate facilities, we found that agreement between the two teams was comparable to agreement between repeated visits by a single team. This result shows that diffuse scattering protocols are transferrable between research groups and synchrotron beamlines, resolving questions about whether these methods were mature enough to be adopted by the wider structural biology community.\n\nBased on our results for Mac1 and previous work [](https://doi.org/10.1107/S2059798324002705) [](https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380), we conclude that obtaining high-quality diffuse scattering data depends on controlling the sample environment, the diffracting volume, the background, and the X-ray dose. Humidity should be controlled during both harvesting and data collection, particularly when multiple crystals are required for complete reciprocal-space coverage. Consistent humidity minimizes crystal-to-crystal variation and is therefore central to reproducibility. Accurate background matching is essential for subtracting scattering from the beamline and the mounting materials: for our high-quality datasets, background collected ~100–300 μm away from the crystal on the sleeve provided a suitable match, whereas background collected at the neck of the Kapton loop did not, likely due to oversubtraction. Low-dose data collection and consistent processing are important for preserving diffuse signal quality and enabling comparisons across beamtimes and beamlines. The signal-to-noise ratio of diffuse scattering datasets can be boosted both by the number of crystals included in a dataset and/or the crystal size. The number of crystals required for a high-quality dataset will depend on their orientations as well as other factors that influence diffuse data quality, but generally more crystals improve the overall quality. In future work, we plan to construct diffuse maps from even larger numbers of crystals to determine how signal quality continues to improve with additional datasets. Larger crystals are generally advantageous because they increase the diffracting volume and therefore the signal-to-noise ratio. For Mac1, 500 μm crystals give exceptionally strong diffuse signal, whereas 100 μm crystals give noticeably weaker signal. Whether the small crystal signal level is sufficient to interpret protein motions remains an open question. We are currently investigating how diffuse signal-to-noise propagates into the models derived from these maps, which will determine what crystal size and signal strength are needed for a given interpretive claim.\n\nAmong the effects observed in the non-ideal datasets, the dose-dependent changes in diffuse scattering were the most surprising and least well understood. In Mac1, increasing dose produced an increase in the anisotropic signal at low resolution and the emergence of a pattern distinct from that of the undamaged sample. We also observed a modest increase in the overall B-factor during scaling, consistent with global radiation damage and loss of resolution. Previous diffuse scattering studies have shown that diffuse halos and overall B-factors can be described by an elastic model of the crystal lattice with thermally excited vibrations [](https://doi.org/10.1038/s41467-023-36734-3). Within this framework, the simplest model of global radiation damage would be a dose-dependent increase in the effective temperature, or a decrease in the stiffness, of the elastic model. In the one-phonon approximation, this model predicts a resolution-dependent rescaling of the anisotropic diffuse pattern as the B-factor increases, analogous to the effect on Bragg intensities [](https://doi.org/10.1107/S1399004715013838). However, the dose-dependent change in diffuse scattering we observe is inconsistent with this simple model: the changes occur predominantly at low resolution, and the intensity pattern evolves with dose rather than simply being rescaled. One possible explanation is that radiation damage is concentrated in microscopic regions or defects, consistent with recent 4D-STEM observations of nanoscale domain reorientation in lysozyme and myoglobin crystals [](https://doi.org/10.1016/j.bpj.2026.08.008). These results suggest that diffuse scattering may provide a novel way to track these more complex radiation-damage mechanisms in protein crystallography. However, given the apparent complexity of these mechanisms, we might expect different crystals to respond differently to X-ray dose. Future work will study these effects systematically across a range of systems beyond Mac1. Importantly, we identify a low-dose regime in which radiation-damage effects are negligible relative to other sources of error.\n\nTogether, these results demonstrate cross-facility diffuse scattering generalizability, offer a statistical framework for evaluating reproducibility, and clarify the experimental conditions for high-quality diffuse data. These are all prerequisites for the next steps: perturbation experiments that examine how the diffuse signal changes in response to factors such as ligand binding or temperature. Mac1 is well suited to such studies given its compatibility with a wide range of ligands, and several perturbation datasets have already been collected (see [online logbooks](https://diffuse.science/logbook/beamtime)). Since this work characterizes the variation between independent measurements, perturbation-induced changes in those datasets can be judged against a known baseline, even when they are subtle. In addition, the high-quality datasets reported here will support advanced modeling approaches, such as separating lattice contributions from internal protein motions using GOODVIBES and 3D-∆PDF analysis [](https://doi.org/10.1038/s41467-023-36734-3). They will also allow models of protein dynamics to be tested directly against experimental diffuse scattering, one of the long-term goals of the sampleworks project [](https://doi.org/10.82153/JKXJ-TW08). More broadly, establishing procedures for trustworthy and accurate measurements will help realize the potential of diffuse scattering to reveal protein motion through X-ray crystallography.\n\n# **Methods**\n\n## **Protein expression, purification, and crystallization**\n\nThe P4$_{3}$ Mac1 construct (residues 3-169) was expressed and purified at UCSF as described previously [](https://doi.org/10.1126/sciadv.abo5083) [](https://doi.org/10.1126/sciadv.abf8711). Briefly, Mac1 with an N-terminal His tag was expressed in E. coli BL21(DE3). Cultures were grown in ZY autoinduction media at 37°C to OD600 = 0.8, then the temperature was lowered to 20°C for expression for an additional 12 hrs. Cells were harvested by centrifugation and stored at -80°C. All chromatography was performed on an ÄKTA Pure FPLC at 4°C. Thawed pellets were resuspended in His load buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, 10 mM imidazole, 5% glycerol, 2 mM βME) with Turbonuclease (Sigma) and lysed by sonication. Clarified lysate was purified on a HisTrap HP column (Cytiva), and Mac1 was eluted with His load buffer supplemented with 300 mM imidazole. The eluate was desalted into TEV buffer and cleaved with TEV protease (1:10 w/w) overnight at 4°C. The reaction was then passed over the HisTrap HP column to separate tagged from un-tagged Mac1. The flowthrough was concentrated and applied to a Superdex 75 16/600 column (Cytiva) pre-equilibrated with storage buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 5% glycerol, 2 mM DTT). Peak fractions were concentrated to 40 mg/mL, flash-frozen, and stored at -80°C. Frozen crystal seed stocks were obtained from previous crystallization trials.\n\nAt UCSF, crystals were grown by sitting-drop vapor diffusion in 96-well 3-drop Swissci plates at ambient temperature (20°C). Drops of 400 nL total volume (200 nL protein, 100 nL seeds, 100 nL reservoir) were equilibrated against reservoir solution containing 100 mM CHES pH 9.5 and 24–35% PEG 3000. Protein concentration was held constant, while the reservoir PEG concentration was varied to control crystal size: 34% PEG 3000 was used to grow the large crystals for the October and November beamtimes, and 28% PEG 3000 to grow the smaller crystals for the April beamtime.\n\nFrozen protein aliquots, seeds, and crystallization reagents were shipped from UCSF to Cornell. At Cornell, crystals were grown from 24-well hanging-drop vapor diffusion trays at ambient temperature (~21°C). Drops of 2 µl protein solution + 1 ul well solution + 1 ul seeds were equilibrated over a reservoir containing 100 mM CHES pH 9.5 and 28-38% PEG 3000. The PEG concentration and seed dilution were varied to optimize for large single crystals. Crystals collected during the September 2025 CHESS beamtime were harvested from a condition with 34% PEG and undiluted seed stock. Crystals collected during the November 2025 CHESS beamtime were harvested from a condition with 30% PEG and seed stock diluted 10-fold in water.\n\n## **Data collection**\n\n::::::div{.info-box}\nRead a comprehensive account of each synchrotron visit in the [online logbook](https://diffuse.science/logbook), generated from [this GitHub repo](https://github.com/prism-science/logbook) (DOI: [10.5281/zenodo.17351307](https://doi.org/10.5281/zenodo.17351307)).\n::::::\n\nHere, we summarize the beamline configuration, sample handling, and data collection procedures used for the four high-quality datasets.\n\nThe East team collected diffraction data at CHESS beamline ID7B2 using an Eiger2 X 16M detector. The ID7B2 beamline is designed for high throughput cryo-crystallography and room temperature serial crystallography of microcrystals, and features a high-flux multilayer monochromator and CRL-focused beam. To adapt this beamline for diffuse scattering, a channel cut Si crystal was inserted to reduce the energy bandwidth from 0.7% defined by the multilayer optics to a nominal 0.014% (Si 111 reflection). A spatially uniform (\"top-hat\") intensity profile was obtained by bypassing the CRL focusing optic and defining the beam shape with slits set to 100 µm x 100 µm. The beam was focused using a toroidal mirror 10 m upstream of the sample, resulting in a nominal divergence of 0.1 mrad (0.006˚) in both the horizontal and vertical directions. A 600 µm ID Mo tube was added upstream of the sample to reduce air scatter and block parasitic diffraction from the slits. Extra metal shielding was also applied to prevent stray upstream scattering and fluorescence from reaching the detector. An X-ray energy of 14 keV was chosen to maximize the amount of high resolution data that could be collected from the well-diffracting Mac1 crystals, while maintaining reasonable detector efficiency, which falls off rapidly above this energy. At the start of each beamtime, the flux was measured using an N$_{2}$-filled ion chamber. The dose rate was controlled by adjusting the exposure time per frame (typically 10 ms) and inserting Al attenuator foils if necessary to reduce the flux.\n\nThe West team collected diffraction data at ALS beamline 8.3.1 using a Pilatus3 S 6M detector and a 100 µm diameter X-ray beam with 2 mrad (0.1˚) horizontal and ~0.35 mrad (0.02˚) vertical divergence. The beam shape was defined by the combination of focusing optics and a round pinhole aperture, and the resulting intensity profile was approximately Gaussian. This profile results in a higher flux density at the beam center than at the beam edges, and an accumulated dose that is typically less spatially uniform compared with a top-hat beam. Beam intensity was controlled using attenuating foils (which adjust the transmission factor) and adjustments to the horizontal focusing optics. An X-ray energy of 14 keV was chosen for data collection during the October 2025 beamtime. During the November 2025 beamtime, additional collimator scatter was observed at this energy, and the energy was lowered to 11.145 keV to avoid this source of background. Because less photon flux is available at higher energies, the dose rate was considerably different for these two beamtimes: approximately 1 kGy/s in October and 2.5 kGy/s in November. This was compensated by reducing the exposure time per 0.1˚ frame from 50 ms in October to 20 ms in November.\n\nSamples were mounted at the beamline directly from crystallization trays. At CHESS, all sample handling was performed in a humidified chamber (MiTeGen SLEEC, ~100% RH). The same method was used at ALS, except in the case of two crystals mounted during the October 2025 beamtime, where a stream of humidified gas was used instead (MiTeGen Watershed, ~90% RH). Further details can be found in the Logbook. Crystals harvested with Kapton loops were placed within plastic capillaries (MiTeGen MicroRT) with well solution added to the tip. To ensure consistent vapor equilibrium, the capillaries were cut to length and filled so that the meniscus was within 2-4 mm of the crystal. The capillary ensheathed crystals were then mounted by hand on the goniometer. To correct for background from the experimental apparatus and sample mounting materials, we followed our established protocol of collecting a 360 degree background dataset for each crystal (1 degree per frame) with the crystal translated out of the beam 100-300 µm along the spindle axis direction [](https://doi.org/10.1038/s41467-020-14933-6) [](https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380). The exposure time for background frames was set to maintain the same incident fluence per degree of rotation for a consistent match. For instance, at CHESS exposures were typically 100 ms per 1˚ for background datasets vs 10 ms per 0.1˚ for the crystal datasets.\n\n## **Data processing**\n\n::::::div{.info-box}\nData, processing, and analysis are recorded using [Datalad](https://www.datalad.org/) datasets, which track data provenance alongside code and link to a public object store for downloading large files.\n\n**Raw X-ray diffraction images** for the four high-quality datasets are available from [this GitHub repo](https://github.com/prism-science/diffuse-data-pub-mac1-east-west) (DOI: [10.5281/zenodo.23487904](https://doi.org/10.5281/zenodo.23487904)), and images for the lesser-quality dataset are available from [this GitHub repo](https://github.com/prism-science/diffuse-data-pub-mac1-suboptimal) (DOI: [10.5281/zenodo.23487927](https://doi.org/10.5281/zenodo.23487927)). The four high-quality datasets have also been deposited in the SBGrid Databank with IDs [1301](https://data.sbgrid.org/dataset/1301/), [1302](https://data.sbgrid.org/dataset/1302), [1303](https://data.sbgrid.org/dataset/1303), and [1304](https://data.sbgrid.org/dataset/1304).\n\n**Data processing scripts and results** (including text logs, mtz files for Bragg data, and nexus-formatted hdf5 files for the diffuse maps) are available from [this GitHub repo](https://github.com/prism-science/diffuse-processing-mac1-east-west) (DOI: [10.5281/zenodo.23487517](https://doi.org/10.5281/zenodo.23487517)).\n\n**Code** to analyze the processed data and reproduce the figures in this article is available from [this GitHub repo](https://github.com/prism-science/diffuse-analysis-mac1-east-west) (DOI: [10.5281/zenodo.23487401](https://doi.org/10.5281/zenodo.23487401)).\n::::::\n\nOur data processing procedure used DIALS [](https://onlinelibrary.wiley.com/doi/10.1002/pro.4224) for standard Bragg data reduction and geometry refinement followed by mdx2 [](https://doi.org/10.1107/S2059798324002705) for diffuse scattering integration, scaling, merging, and visualization. The overall multi-crystal workflow, diffuse data processing principles, and their implementation in mdx2 have been described previously [](https://doi.org/10.1107/S2059798324002705). Briefly, mdx2 assigns each pixel on the detector to a location in reciprocal space using the refined geometry from DIALS, and accumulates photon counts on a three-dimensional grid. The grid is oversampled relative to the reciprocal lattice in order to capture the strong diffuse intensity variations near the Bragg peaks (\"halos\"). After integrating data from each crystal separately, the datasets were corrected for background scattering and scaled collectively with empirically refined corrections for residual isotropic background, anisotropic absorption, detector inhomogeneity, and variations in illuminated volume. The scaled intensities are then symmetry-averaged and merged.\n\nFor Mac1, a moderate degree of oversampling was chosen (2x2x4, or voxels having dimensions of a*/2, b*/2, c*/4), sufficient to separate the strong halo intensities from the more slowly varying cloudy patterns, while maintaining a relatively high signal-to-noise within each voxel for a more statistically robust comparison between datasets. For visualization purposes, the merged intensities are typically symmetry-expanded (the asymmetric unit for Laue group 4/m occupies 1/8 of reciprocal space, chosen as the positive octant).\n\nDiffuse scattering from Mac1 has a characteristic peak in the isotropic background at ~3.3 Å resolution. For each dataset, we estimated the isotropic component by computing a trimmed mean (excluding the top and bottom 10% of intensities) within narrow resolution shells (width 0.01 Å$^{-1}$). To visualize the more subtle pattern of variations on top of the strong isotropic scattering, we also compute an \"anisotropic\" scattering component by subtracting the isotropic scattering interpolated at each voxel.\n\nThe intensity scale from mdx2 is in units of photons per second per solid angle. We have previously developed procedures for placing the intensity on an absolute scale of electron units that can be compared quantitatively with simulations [](https://doi.org/10.1038/s41467-020-14933-6). However, for assessing replication, it is sufficient to normalize the data on an arbitrary, relative scale. We chose to normalize each dataset by the peak isotropic value. Thus, on this normalized scale, most of the measured intensities range from 0-1, with bright features somewhat exceeding the peak in the water ring.\n\nTo compute CC$_{1/2}$, the unmerged intensities from each dataset were randomly partitioned into two groups and merged separately using mdx2. Then, the common isotropic component was subtracted, and Pearson correlation coefficients between the anisotropic maps were computed within narrow resolution shells (width 0.01 Å$^{-1}$). In this work, CCs are always computed using the anisotropic pattern, rather than the total scattering, because the isotropic component can vary significantly within certain resolution bins and therefore inflate the CC value [](https://doi.org/10.1073/pnas.1416744111).\n\n## **Derivation of CC_lim for dataset comparisons**\n\nWe developed a statistical measure, CC$_{\\text{lim}}$, to understand the maximum expected correlation coefficient of two independent datasets of the same signal. We present the derivation below given dataset $a=\\ \\tau \\ +\\ {\\epsilon }_{A}$ and $b=\\ \\tau \\ +\\ {\\epsilon }_{B}$, where $\\tau$ is the signal and $\\epsilon$ is the associated noise. The correlation will be\n\n$$corr(a,b)\\ =\\frac{cov(a,b)}{\\sqrt{var(a)var(b)}}{\\ }$$\n\nTaking the numerator:\n\n$$cov(a,b)=cov(\\tau \\ +\\ {\\epsilon }_{A},\\tau \\ +\\ {\\epsilon }_{B})=cov(\\tau ,\\tau )+cov(\\tau {,\\epsilon }_{B})+cov({\\epsilon }_{A},\\tau )+cov({\\epsilon }_{A}{,\\epsilon }_{B})={\\sigma }_{\\tau }^{2}$$\n\nsince the noise is independent from each other and from the signal. Taking the denominator, for each dataset we get:\n\n$$var(a)=var(\\tau \\ +\\ {\\epsilon }_{A})=var(\\tau )+var(\\ {\\epsilon }_{A})+2cov(\\tau \\ ,{\\epsilon }_{A})={\\sigma }_{\\tau }^{2}+{\\sigma }_{{\\epsilon }_{A}}^{2}$$\n\nTherefore, we are left with:\n\n$$corr(a,b)=\\frac{{\\sigma }_{\\tau }^{2}}{\\sqrt{({\\sigma }_{\\tau }^{2}+{\\sigma }_{{\\epsilon }_{A}}^{2}){(\\sigma }_{\\tau }^{2}+{\\sigma }_{{\\epsilon }_{B}}^{2})}}=\\frac{{\\sigma }_{\\tau }^{}}{\\sqrt{{\\sigma }_{\\tau }^{2}+{\\sigma }_{{\\epsilon }_{A}}^{2}}}\\frac{{\\sigma }_{\\tau }^{}}{\\sqrt{{\\sigma }_{\\tau }^{2}+{\\sigma }_{{\\epsilon }_{B}}^{2}}}={CC}_{A}^{*}{CC}_{B}^{*}$$\n\n,which is simply the product of the CC* of each dataset [](https://doi.org/10.1126/science.1218231). This allows us to understand how similar two datasets are and if they deviate beyond their noise level. Specifically, if CC$_{\\text{pair}}$ between two datasets is less than CC$_{\\text{lim}}$, then they vary beyond the noise.\n\n::::::div{.bordered.lead data-internal=\"acknowledgements\"}\n### **Acknowledgements**\n\n* This work was supported by Radial, a division of Astera Institute.\n* Beamline 8.3.1 at the Advanced Light Source is operated by the University of California San Francisco with generous support from the National Institutes of Health (P30 GM124169 for user support), and the Integrated Diffraction Analysis Technologies program of the US Department of Energy Office of Biological and Environmental Research. The Advanced Light Source (Berkeley, CA) is a national user facility operated by Lawrence Berkeley National Laboratory on behalf of the US Department of Energy under contract number DE-AC02-05CH11231, Office of Basic Energy Sciences.\n* Beamline ID7B2 at CHESS is part of the Center for High-Energy X-ray Sciences (CHEXS), which is supported by the National Science Foundation (BIO, ENG and MPS Directorates) under award DMR-2342336, and the Macromolecular Diffraction at CHESS (MacCHESS) facility, which is supported by award P30GM124166 from the National Institute of General Medical Sciences and the National Institutes of Health. Beamtime for this project was supported by funds from Cornell University.\n* Raw data and results are shared publicly on the Open Storage Network (OSN) via project allocation BIO250275 from ACCESS[](https://doi.org/10.1145/3569951.3597559), supported by U.S. National Science Foundation (NSF) grants #2138259, #2138286, #2138307, #2137603, and #2138296.\n::::::\n\n::::::bibtex\n@article{abramson2024, author = {Abramson, Josh and Adler, Jonas and Dunger, Jack and Evans, Richard and Green, Tim and Pritzel, Alexander and Ronneberger, Olaf and Willmore, Lindsay and Ballard, Andrew J. and Bambrick, Joshua and Bodenstein, Sebastian W. and Evans, David A. and Hung, Chia-Chun and O’Neill, Michael and Reiman, David and Tunyasuvunakool, Kathryn and Wu, Zachary and Žemgulytė, Akvilė and Arvaniti, Eirini and Beattie, Charles and Bertolli, Ottavia and Bridgland, Alex and Cherepanov, Alexey and Congreve, Miles and Cowen-Rivers, Alexander I. and Cowie, Andrew and Figurnov, Michael and Fuchs, Fabian B. and Gladman, Hannah and Jain, Rishub and Khan, Yousuf A. and Low, Caroline M. R. and Perlin, Kuba and Potapenko, Anna and Savy, Pascal and Singh, Sukhdeep and Stecula, Adrian and Thillaisundaram, Ashok and Tong, Catherine and Yakneen, Sergei and Zhong, Ellen D. and Zielinski, Michal and Žídek, Augustin and Bapst, Victor and Kohli, Pushmeet and Jaderberg, Max and Hassabis, Demis and Jumper, John M.}, title = {Accurate structure prediction of biomolecular interactions with AlphaFold 3}, journal = {Nature}, volume = {630}, number = {8016}, pages = {493–500}, publisher = {Springer Science and Business Media LLC}, year = {2024}, month = {May}, doi = {10.1038/s41586-024-07487-w} }\n\n@article{bowman2024, author = {Bowman, Gregory R.}, title = {AlphaFold and Protein Folding: Not Dead Yet! The Frontier Is Conformational Ensembles}, journal = {Annual Review of Biomedical Data Science}, volume = {7}, number = {1}, pages = {51–57}, publisher = {Annual Reviews}, year = {2024}, month = {Aug}, doi = {10.1146/annurev-biodatasci-102423-011435} }\n\n@article{burley2022, author = {Burley, Stephen K. and Berman, Helen M. and Duarte, Jose M. and Feng, Zukang and Flatt, Justin W. and Hudson, Brian P. and Lowe, Robert and Peisach, Ezra and Piehl, Dennis W. and Rose, Yana and Sali, Andrej and Sekharan, Monica and Shao, Chenghua and Vallat, Brinda and Voigt, Maria and Westbrook, John D. and Young, Jasmine Y. and Zardecki, Christine}, title = {Protein Data Bank: A Comprehensive Review of 3D Structure Holdings and Worldwide Utilization by Researchers, Educators, and Students}, journal = {Biomolecules}, volume = {12}, number = {10}, pages = {1425}, publisher = {MDPI AG}, year = {2022}, month = {Oct}, doi = {10.3390/biom12101425} }\n\n@misc{chrispens2026, author = {Chrispens, Karson and Collins, Marcus and Fraser, James S. and Mai, Doris and van den Bedem, Henry and Wankowicz, Stephanie A}, title = {sampleworks: A Modular Platform for Experimentally Guided Biomolecular Ensemble Generation}, publisher = {Radial}, year = {2026}, doi = {10.82153/JKXJ-TW08} }\n\n@article{correy2022, author = {Correy, Galen J. and Kneller, Daniel W. and Phillips, Gwyndalyn and Pant, Swati and Russi, Silvia and Cohen, Aina E. and Meigs, George and Holton, James M. and Gahbauer, Stefan and Thompson, Michael C. and Ashworth, Alan and Coates, Leighton and Kovalevsky, Andrey and Meilleur, Flora and Fraser, James S.}, title = {The mechanisms of catalysis and ligand binding for the SARS-CoV-2 NSP3 macrodomain from neutron and x-ray diffraction at room temperature}, journal = {Science Advances}, volume = {8}, number = {21}, publisher = {American Association for the Advancement of Science (AAAS)}, year = {2022}, month = {May}, doi = {10.1126/sciadv.abo5083} }\n\n@article{jackson2007, author = {Jackson, Colin J. and Carr, Paul D. and Liu, Jian-Wei and Watt, Stephen J. and Beck, Jennifer L. and Ollis, David L.}, title = {The Structure and Function of a Novel Glycerophosphodiesterase from Enterobacter aerogenes}, journal = {Journal of Molecular Biology}, volume = {367}, number = {4}, pages = {1047–1062}, publisher = {Elsevier BV}, year = {2007}, month = {Apr}, doi = {10.1016/j.jmb.2007.01.032} }\n\n@article{karplus2012, author = {Karplus, P. Andrew and Diederichs, Kay}, title = {Linking Crystallographic Model and Data Quality}, journal = {Science}, volume = {336}, number = {6084}, pages = {1030–1033}, publisher = {American Association for the Advancement of Science (AAAS)}, year = {2012}, month = {May}, doi = {10.1126/science.1218231} }\n\n@article{lane2023, author = {Lane, Thomas J.}, title = {Protein structure prediction has reached the single-structure frontier}, journal = {Nature Methods}, volume = {20}, number = {2}, pages = {170–173}, publisher = {Springer Science and Business Media LLC}, year = {2023}, month = {Jan}, doi = {10.1038/s41592-022-01760-4} }\n\n@article{meinhold2007, author = {Meinhold, Lars and Smith, Jeremy C.}, title = {Protein dynamics from X‐ray crystallography: Anisotropic, global motion in diffuse scattering patterns}, journal = {Proteins: Structure, Function, and Bioinformatics}, volume = {66}, number = {4}, pages = {941–953}, publisher = {Wiley}, year = {2007}, month = {Feb}, doi = {10.1002/prot.21246} }\n\n@article{meisburger2024, author = {Meisburger, Steve P. and Ando, Nozomi}, title = {Scaling and merging macromolecular diffuse scattering with mdx2}, journal = {Acta Crystallographica Section D Structural Biology}, volume = {80}, number = {5}, pages = {299–313}, publisher = {International Union of Crystallography (IUCr)}, year = {2024}, month = {Apr}, doi = {10.1107/S2059798324002705} }\n\n@article{meisburger2017, author = {Meisburger, Steve P. and Thomas, William C. and Watkins, Maxwell B. and Ando, Nozomi}, title = {X-ray Scattering Studies of Protein Structural Dynamics}, journal = {Chemical Reviews}, volume = {117}, number = {12}, pages = {7615–7672}, publisher = {American Chemical Society (ACS)}, year = {2017}, month = {May}, doi = {10.1021/acs.chemrev.6b00790} }\n\n@article{meisburger2020, author = {Meisburger, Steve P. and Case, David A. and Ando, Nozomi}, title = {Diffuse X-ray scattering from correlated motions in a protein crystal}, journal = {Nature Communications}, volume = {11}, number = {1}, publisher = {Springer Science and Business Media LLC}, year = {2020}, month = {Mar}, doi = {10.1038/s41467-020-14933-6} }\n\n@article{meisburger2023, author = {Meisburger, Steve P. and Case, David A. and Ando, Nozomi}, title = {Robust total X-ray scattering workflow to study correlated motion of proteins in crystals}, journal = {Nature Communications}, volume = {14}, number = {1}, publisher = {Springer Science and Business Media LLC}, year = {2023}, month = {Mar}, doi = {10.1038/s41467-023-36734-3} }\n\n@article{nia2026, author = {Nia, Shervin S. and Saha, Ambarneil and Cho, Jung and Zhou, Z. Hong and Ercius, Peter and Mecklenburg, Matthew}, title = {Snapshots of internal protein crystal architecture at the nanoscale}, journal = {Biophysical Journal}, publisher = {Elsevier BV}, year = {2026}, month = {Aug}, doi = {10.1016/j.bpj.2026.08.008} }\n\n@article{peck2018, author = {Peck, Ariana and Poitevin, Frédéric and Lane, Thomas J.}, title = {Intermolecular correlations are necessary to explain diffuse scattering from protein crystals}, journal = {IUCrJ}, volume = {5}, number = {2}, pages = {211–222}, publisher = {International Union of Crystallography (IUCr)}, year = {2018}, month = {Feb}, doi = {10.1107/S2052252518001124} }\n\n@incollection{pei20231, title = {Chapter One - Introduction to diffuse scattering and data collection}, editor = {Nozomi Ando}, series = {Methods in Enzymology}, publisher = {Academic Press}, volume = {688}, pages = {1-42}, year = {2023},booktitle = {Crystallography of Protein Dynamics}, issn = {0076-6879}, doi = {10.1016/bs.mie.2023.07.007}, url = {https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380}, author = {{Pei X} and {Bhatt N} and {Wang H} and {Ando N} and {Meisburger SP}}, abstract = {A long-standing goal in X-ray crystallography has been to extract information about the collective motions of proteins from diffuse scattering: the weak, textured signal that is found in the background of diffraction images. In the past few years, the field of macromolecular diffuse scattering has seen dramatic progress, and many of the past challenges in measurement and interpretation are now considered tractable. However, the concept of diffuse scattering is still new to many researchers, and a general set of procedures needed to collect a high-quality dataset has never been described in detail. Here, we provide the first guidelines for performing diffuse scattering experiments, which can be performed at any macromolecular crystallography beamline that supports room-temperature studies with a direct detector. We begin with a brief introduction to the theory of diffuse scattering and then walk the reader through the decision-making processes involved in preparing for and conducting a successful diffuse scattering experiment. Finally, we define quality metrics and describe ways to assess data quality both at the beamline and at home. Data obtained in this way can be processed independently by crystallographic software and diffuse scattering software to produce both a crystal structure, which represents the average atomic coordinates, and a three-dimensional diffuse scattering map that can then be interpreted in terms of models for protein motions.} }\n\n@article{polikanov2015, author = {Polikanov, Yury S. and Moore, Peter B.}, title = {Acoustic vibrations contribute to the diffuse scatter produced by ribosome crystals}, journal = {Acta Crystallographica Section D Biological Crystallography}, volume = {71}, number = {10}, pages = {2021–2031}, publisher = {International Union of Crystallography (IUCr)}, year = {2015}, month = {Sept}, doi = {10.1107/S1399004715013838} }\n\n@article{poon2007, author = {Poon, Billy K. and Chen, Xiaorui and Lu, Mingyang and Vyas, Nand K. and Quiocho, Florante A. and Wang, Qinghua and Ma, Jianpeng}, title = {Normal mode refinement of anisotropic thermal parameters for a supramolecular complex at 3.42-Å crystallographic resolution}, journal = {Proceedings of the National Academy of Sciences}, volume = {104}, number = {19}, pages = {7869–7874}, publisher = {National Academy of Sciences}, year = {2007}, month = {May}, doi = {10.1073/pnas.0701204104} }\n\n@article{schuller2021, author = {Schuller, Marion and Correy, Galen J. and Gahbauer, Stefan and Fearon, Daren and Wu, Taiasean and Díaz, Roberto Efraín and Young, Iris D. and Carvalho Martins, Luan and Smith, Dominique H. and Schulze-Gahmen, Ursula and Owens, Tristan W. and Deshpande, Ishan and Merz, Gregory E. and Thwin, Aye C. and Biel, Justin T. and Peters, Jessica K. and Moritz, Michelle and Herrera, Nadia and Kratochvil, Huong T. and Aimon, Anthony and Bennett, James M. and Brandao Neto, Jose and Cohen, Aina E. and Dias, Alexandre and Douangamath, Alice and Dunnett, Louise and Fedorov, Oleg and Ferla, Matteo P. and Fuchs, Martin R. and Gorrie-Stone, Tyler J. and Holton, James M. and Johnson, Michael G. and Krojer, Tobias and Meigs, George and Powell, Ailsa J. and Rack, Johannes Gregor Matthias and Rangel, Victor L. and Russi, Silvia and Skyner, Rachael E. and Smith, Clyde A. and Soares, Alexei S. and Wierman, Jennifer L. and Zhu, Kang and O’Brien, Peter and Jura, Natalia and Ashworth, Alan and Irwin, John J. and Thompson, Michael C. and Gestwicki, Jason E. and von Delft, Frank and Shoichet, Brian K. and Fraser, James S. and Ahel, Ivan}, title = {Fragment binding to the Nsp3 macrodomain of SARS-CoV-2 identified through crystallographic screening and computational docking}, journal = {Science Advances}, volume = {7}, number = {16}, publisher = {American Association for the Advancement of Science (AAAS)}, year = {2021}, month = {Apr}, doi = {10.1126/sciadv.abf8711} }\n\n@inbook{snell2003, author = {Snell, Edward H and Bellamy, Henry D and Borgstahl, Gloria E.O}, title = {Macromolecular Crystal Quality}, booktitle = {Macromolecular Crystallography, Part C}, pages = {268–288}, publisher = {Elsevier}, year = {2003}, doi = {10.1016/S0076-6879(03)68015-8} }\n\n@article{su2021, author = {Su, Zhen and Dasgupta, Medhanjali and Poitevin, Frédéric and Mathews, Irimpan I. and van den Bedem, Henry and Wall, Michael E. and Yoon, Chun Hong and Wilson, Mark A.}, title = {Reproducibility of protein x-ray diffuse scattering and potential utility for modeling atomic displacement parameters}, journal = {Structural Dynamics}, volume = {8}, number = {4}, publisher = {AIP Publishing}, year = {2021}, month = {July}, doi = {10.1063/4.0000087} }\n\n@article{vandenbedem2013, author = {van den Bedem, Henry and Bhabha, Gira and Yang, Kun and Wright, Peter E and Fraser, James S}, title = {Automated identification of functional dynamic contact networks from X-ray crystallography}, journal = {Nature Methods}, volume = {10}, number = {9}, pages = {896–902}, publisher = {Springer Science and Business Media LLC}, year = {2013}, month = {Aug}, doi = {10.1038/nmeth.2592} }\n\n@article{wall1997, author = {Wall, Michael E. and Ealick, Steven E. and Gruner, Sol M.}, title = {Three-dimensional diffuse x-ray scattering from crystals of Staphylococcal nuclease}, journal = {Proceedings of the National Academy of Sciences}, volume = {94}, number = {12}, pages = {6180–6184}, publisher = {National Academy of Sciences}, year = {1997}, month = {June}, doi = {10.1073/pnas.94.12.6180} }\n\n@article{wall2014, author = {Wall, Michael E. and Van Benschoten, Andrew H. and Sauter, Nicholas K. and Adams, Paul D. and Fraser, James S. and Terwilliger, Thomas C.}, title = {Conformational dynamics of a crystalline protein from microsecond-scale molecular dynamics simulations and diffuse X-ray scattering}, journal = {Proceedings of the National Academy of Sciences}, volume = {111}, number = {50}, pages = {17887–17892}, publisher = {National Academy of Sciences}, year = {2014}, month = {Dec}, doi = {10.1073/pnas.1416744111} }\n\n@article{wall2018, author = {Wall, Michael E and Wolff, Alexander M and Fraser, James S}, title = {Bringing diffuse X-ray scattering into focus}, journal = {Current Opinion in Structural Biology}, volume = {50}, pages = {109–116}, publisher = {Elsevier BV}, year = {2018}, month = {June}, doi = {10.1016/j.sbi.2018.01.009} }\n\n@article{wankowicz2026, author = {Wankowicz, Stephanie A. and Bonomi, Massimiliano}, title = {From possibility to precision in macromolecular ensemble prediction}, journal = {Nature Methods}, volume = {23}, number = {6}, pages = {1100–1108}, publisher = {Springer Science and Business Media LLC}, year = {2026}, month = {May}, doi = {10.1038/s41592-026-03084-z} }\n\n@misc{wankowicz2026a, author = {Wankowicz, Stephanie A}, title = {Recovering Conformational Heterogeneity from the Protein Data Bank at Scale}, publisher = {Radial}, year = {2026}, doi = {10.82153/PFF0-CK46} }\n\n@article{winter2022, author = {{Winter G} and {Beilsten‐Edmands J} and {Devenish N} and {Gerstel M} and {Gildea RJ} and {McDonagh D} and {Pascal E} and {Waterman DG} and {Williams BH} and {Evans G}}, title = {DIALS as a toolkit}, journal = {Protein Science}, volume = {31}, number = {1}, pages = {232–250}, publisher = {Wiley}, year = {2022}, month = {Nov}, doi = {10.1002/pro.4224}, url = {https://onlinelibrary.wiley.com/doi/10.1002/pro.4224} }\n\n@article{xu2021, author = {Xu, Da and Meisburger, Steve P. and Ando, Nozomi}, title = {Correlated Motions in Structural Biology}, journal = {Biochemistry}, volume = {60}, number = {30}, pages = {2331–2340}, publisher = {American Chemical Society (ACS)}, year = {2021}, month = {July}, doi = {10.1021/acs.biochem.1c00420} }\n::::::\n\n","contributors":[{"user_id":5567,"role":"Conceptualization","first_name":"Kara","last_name":"Zielinski","avatar":null,"in_byline":1,"priority":1,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5567,"role":"Investigation","first_name":"Kara","last_name":"Zielinski","avatar":null,"in_byline":1,"priority":1,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5567,"role":"Methodology","first_name":"Kara","last_name":"Zielinski","avatar":null,"in_byline":1,"priority":1,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5567,"role":"Writing","first_name":"Kara","last_name":"Zielinski","avatar":null,"in_byline":1,"priority":1,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5367,"role":"Conceptualization","first_name":"Steve","last_name":"Meisburger","avatar":"avatars/users/user_5367_1790909397411.jpg","in_byline":1,"priority":2,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null},{"org_id":14,"name":"Cornell High Energy Synchrotron Source","slug":"chess-cornell","ror_id":"032bgnz47","avatar":null}]},{"user_id":5367,"role":"Investigation","first_name":"Steve","last_name":"Meisburger","avatar":"avatars/users/user_5367_1790909397411.jpg","in_byline":1,"priority":2,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null},{"org_id":14,"name":"Cornell High Energy Synchrotron Source","slug":"chess-cornell","ror_id":"032bgnz47","avatar":null}]},{"user_id":5367,"role":"Methodology","first_name":"Steve","last_name":"Meisburger","avatar":"avatars/users/user_5367_1790909397411.jpg","in_byline":1,"priority":2,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null},{"org_id":14,"name":"Cornell High Energy Synchrotron Source","slug":"chess-cornell","ror_id":"032bgnz47","avatar":null}]},{"user_id":5367,"role":"Software","first_name":"Steve","last_name":"Meisburger","avatar":"avatars/users/user_5367_1790909397411.jpg","in_byline":1,"priority":2,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null},{"org_id":14,"name":"Cornell High Energy Synchrotron Source","slug":"chess-cornell","ror_id":"032bgnz47","avatar":null}]},{"user_id":5367,"role":"Writing","first_name":"Steve","last_name":"Meisburger","avatar":"avatars/users/user_5367_1790909397411.jpg","in_byline":1,"priority":2,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null},{"org_id":14,"name":"Cornell High Energy Synchrotron Source","slug":"chess-cornell","ror_id":"032bgnz47","avatar":null}]},{"user_id":5632,"role":"Investigation","first_name":"Katherine","last_name":"Lu","avatar":null,"in_byline":1,"priority":3,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null}]},{"user_id":5572,"role":"Investigation","first_name":"Violla","last_name":"Bassim","avatar":null,"in_byline":1,"priority":4,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5252,"role":"Investigation","first_name":"James","last_name":"Holton","avatar":null,"in_byline":1,"priority":5,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null},{"org_id":11,"name":"SLAC National Accelerator Laboratory","slug":"slac","ror_id":"05gzmn429","avatar":null}]},{"user_id":4928,"role":"Visualization","first_name":"Andrew A","last_name":"Burnim","avatar":null,"in_byline":0,"priority":6,"affiliations":[{"org_id":7,"name":"Radial","slug":"radial","ror_id":"050rbg919","avatar":"avatars/orgs/org_7_1780396782.jpg"}]},{"user_id":5633,"role":"Data curation","first_name":"Ellaine","last_name":"Chou","avatar":null,"in_byline":0,"priority":7,"affiliations":[{"org_id":7,"name":"Radial","slug":"radial","ror_id":"050rbg919","avatar":"avatars/orgs/org_7_1780396782.jpg"}]},{"user_id":5246,"role":"Conceptualization","first_name":"James S.","last_name":"Fraser","avatar":null,"in_byline":1,"priority":8,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5246,"role":"Project administration","first_name":"James S.","last_name":"Fraser","avatar":null,"in_byline":1,"priority":8,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5246,"role":"Supervision","first_name":"James S.","last_name":"Fraser","avatar":null,"in_byline":1,"priority":8,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5246,"role":"Writing","first_name":"James S.","last_name":"Fraser","avatar":null,"in_byline":1,"priority":8,"affiliations":[{"org_id":8,"name":"University of California, San Francisco","slug":"ucsf","ror_id":"043mz5j54","avatar":null}]},{"user_id":5634,"role":"Conceptualization","first_name":"Nozomi","last_name":"Ando","avatar":null,"in_byline":1,"priority":9,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null}]},{"user_id":5634,"role":"Project administration","first_name":"Nozomi","last_name":"Ando","avatar":null,"in_byline":1,"priority":9,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null}]},{"user_id":5634,"role":"Supervision","first_name":"Nozomi","last_name":"Ando","avatar":null,"in_byline":1,"priority":9,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null}]},{"user_id":5634,"role":"Writing","first_name":"Nozomi","last_name":"Ando","avatar":null,"in_byline":1,"priority":9,"affiliations":[{"org_id":13,"name":"Cornell University","slug":"cornell","ror_id":"05bnh6r87","avatar":null}]}],"created_at":"2026-10-20T19:47:36.821Z","doi":"10.82153/hrpy-t681","version_doi":"10.82153/hrpy-t681.2","feedback_form_embed_src":null,"id":187,"license":"CC BY","content":{"format":"gdoc"},"linked_assets":[{"type":"Code + data","url":"https://github.com/prism-science/diffuse-analysis-mac1-east-west","name":"Analysis of diffuse scattering data for apo Mac1"},{"type":"Code + data","url":"https://github.com/prism-science/diffuse-processing-mac1-east-west","name":"Diffuse scattering data processing for apo Mac1"},{"type":"Data","url":"https://github.com/prism-science/diffuse-data-pub-mac1-east-west","name":"Raw X-ray diffraction data for apo Mac1"},{"type":"Data","url":"https://github.com/prism-science/diffuse-data-pub-mac1-suboptimal","name":"Raw X-ray diffraction data for suboptimal apo Mac1 datasets"},{"type":"Material","url":"https://diffuse.science/logbook/beamtime/","name":"Logbook"}],"mediaAssets":[{"id":"01M52NS32RW47HMCTED578FWV3","byteSize":100946,"mimeType":"image/png","kind":"image","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_6676637b_3c94292b03c0","width":696,"height":361,"durationMs":null,"rotation":null},{"id":"01M52NS3DY5VF1B0HNHHX47CDE","byteSize":8600806,"mimeType":"video/mp4","kind":"video","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_73562a7f_c753017f8ced","width":838,"height":800,"durationMs":32100,"rotation":0},{"id":"01M52NS3MXH3KAP583Q5NE4698","byteSize":7227272,"mimeType":"image/png","kind":"image","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_b0626931_ac57d438b6f8","width":2853,"height":3216,"durationMs":null,"rotation":null},{"id":"01M52NS4S1BPXE0HW3DJSR1WJN","byteSize":4837537,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_843e074a_21b4f8d5674e","width":null,"height":null,"durationMs":null,"rotation":null},{"id":"01M52NS5HD070P71BSK0XE0EYR","byteSize":16697495,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_f0783a32_0f3969943b4f","width":null,"height":null,"durationMs":null,"rotation":null},{"id":"01M52NS5RE4PBDX7BXSN4YKNHY","byteSize":4888808,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_8bc6b587_d41b0f09519e","width":null,"height":null,"durationMs":null,"rotation":null},{"id":"01M52NS6W45DMQZFVWZG5RSAHZ","byteSize":12042114,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_67edc2dc_40cfd2e43500","width":null,"height":null,"durationMs":null,"rotation":null},{"id":"01M52NSPBNG563TD3YGZ794ERP","byteSize":28455469,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_0aa91147","width":null,"height":null,"durationMs":null,"rotation":null},{"id":"01M52SQDRR2PK6N5S154GM5VFP","byteSize":789962,"mimeType":"image/png","kind":"image","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_5418fbcd_6072efb90f0e","width":5409,"height":2628,"durationMs":null,"rotation":null},{"id":"01M5A9T7HZD51XM9BWCVBGS0WH","byteSize":28455469,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_0aa91147_11dab49db90a","width":null,"height":null,"durationMs":null,"rotation":null},{"id":"01M5D7HFTPJ9TP3R6XY3AVCKRT","byteSize":12042114,"mimeType":"text/html","kind":"file","publicUrl":"https://thestacks-01.s3.amazonaws.com/publications/01M3MXMRFR5HA8VNDRS7TJPYJ2/media_67edc2dc","width":null,"height":null,"durationMs":null,"rotation":null}],"orgs":[{"org_id_int":7,"priority":null,"slug":"radial","name":"Radial","avatar":"avatars/orgs/org_7_1780396782.jpg"}],"pdf_url":null,"publication_ulid":"01M3MXMRFR5HA8VNDRS7TJPYJ2","publisher":"The Stacks","references":[{"url":"https://doi.org/10.1146/annurev-biodatasci-102423-011435","text":"Bowman GR. (2024). AlphaFold and Protein Folding: Not Dead Yet! The Frontier Is Conformational Ensembles."},{"url":"https://doi.org/10.1038/s41592-022-01760-4","text":"Lane TJ. (2023). Protein structure prediction has reached the single-structure frontier."},{"url":"https://doi.org/10.1021/acs.biochem.1c00420","text":"Xu D, Meisburger SP, Ando N. (2021). Correlated Motions in Structural Biology."},{"url":"https://doi.org/10.1038/s41586-024-07487-w","text":"Abramson J, Adler J, Dunger J, Evans R, Green T, Pritzel A, Ronneberger O, Willmore L, Ballard AJ, Bambrick J, Bodenstein SW, Evans DA, Hung C-C, O’Neill M, Reiman D, Tunyasuvunakool K, Wu Z, Žemgulytė A, Arvaniti E, Beattie C, Bertolli O, Bridgland A, Cherepanov A, Congreve M, Cowen-Rivers AI, Cowie A, Figurnov M, Fuchs FB, Gladman H, Jain R, Khan YA, Low CMR, Perlin K, Potapenko A, Savy P, Singh S, Stecula A, Thillaisundaram A, Tong C, Yakneen S, Zhong ED, Zielinski M, Žídek A, Bapst V, Kohli P, Jaderberg M, Hassabis D, Jumper JM. (2024). Accurate structure prediction of biomolecular interactions with AlphaFold 3."},{"url":"https://doi.org/10.82153/jkxj-tw08","text":"Chrispens K, Collins M, Fraser JS, Mai D, van den Bedem H, Wankowicz SA. (2026). _sampleworks:_ A Modular Platform for Experimentally Guided Biomolecular Ensemble Generation."},{"url":"https://doi.org/10.1038/s41592-026-03084-z","text":"Wankowicz SA, Bonomi M. (2026). From possibility to precision in macromolecular ensemble prediction."},{"url":"https://doi.org/10.3390/biom12101425","text":"Burley SK, Berman HM, Duarte JM, Feng Z, Flatt JW, Hudson BP, Lowe R, Peisach E, Piehl DW, Rose Y, Sali A, Sekharan M, Shao C, Vallat B, Voigt M, Westbrook JD, Young JY, Zardecki C. (2022). Protein Data Bank: A Comprehensive Review of 3D Structure Holdings and Worldwide Utilization by Researchers, Educators, and Students."},{"url":"https://doi.org/10.1016/j.jmb.2007.01.032","text":"Jackson CJ, Carr PD, Liu J, Watt SJ, Beck JL, Ollis DL. (2007). The Structure and Function of a Novel Glycerophosphodiesterase from Enterobacter aerogenes."},{"url":"https://doi.org/10.1073/pnas.0701204104","text":"Poon BK, Chen X, Lu M, Vyas NK, Quiocho FA, Wang Q, Ma J. (2007). Normal mode refinement of anisotropic thermal parameters for a supramolecular complex at 3.42-Å crystallographic resolution."},{"url":"https://doi.org/10.1038/nmeth.2592","text":"van den Bedem H, Bhabha G, Yang K, Wright PE, Fraser JS. (2013). Automated identification of functional dynamic contact networks from X-ray crystallography."},{"url":"https://doi.org/10.82153/pff0-ck46","text":"Wankowicz SA. (2026). Recovering Conformational Heterogeneity from the Protein Data Bank at Scale."},{"url":"https://doi.org/10.1038/s41467-020-14933-6","text":"Meisburger SP, Case DA, Ando N. (2020). Diffuse X-ray scattering from correlated motions in a protein crystal."},{"url":"https://doi.org/10.1107/s2052252518001124","text":"Peck A, Poitevin F, Lane TJ. (2018). Intermolecular correlations are necessary to explain diffuse scattering from protein crystals."},{"url":"https://doi.org/10.1107/s1399004715013838","text":"Polikanov YS, Moore PB. (2015). Acoustic vibrations contribute to the diffuse scatter produced by ribosome crystals."},{"url":"https://doi.org/10.1021/acs.chemrev.6b00790","text":"Meisburger SP, Thomas WC, Watkins MB, Ando N. (2017). X-ray Scattering Studies of Protein Structural Dynamics."},{"url":"https://doi.org/10.1073/pnas.94.12.6180","text":"Wall ME, Ealick SE, Gruner SM. (1997). Three-dimensional diffuse x-ray scattering from crystals of Staphylococcal nuclease."},{"url":"https://doi.org/10.1073/pnas.1416744111","text":"Wall ME, Van Benschoten AH, Sauter NK, Adams PD, Fraser JS, Terwilliger TC. (2014). Conformational dynamics of a crystalline protein from microsecond-scale molecular dynamics simulations and diffuse X-ray scattering."},{"url":"https://doi.org/10.1063/4.0000087","text":"Su Z, Dasgupta M, Poitevin F, Mathews II, van den Bedem H, Wall ME, Yoon CH, Wilson MA. (2021). Reproducibility of protein x-ray diffuse scattering and potential utility for modeling atomic displacement parameters."},{"url":"https://doi.org/10.1107/s2059798324002705","text":"Meisburger SP, Ando N. (2024). Scaling and merging macromolecular diffuse scattering with mdx 2."},{"url":"https://doi.org/10.1016/j.sbi.2018.01.009","text":"Wall ME, Wolff AM, Fraser JS. (2018). Bringing diffuse X-ray scattering into focus."},{"url":"https://www.sciencedirect.com/science/chapter/bookseries/abs/pii/S0076687923002380","text":"Pei X, Bhatt N, Wang H, Ando N, Meisburger SP. (2023). Chapter One - Introduction to diffuse scattering and data collection."},{"url":"https://doi.org/10.1126/sciadv.abf8711","text":"Schuller M, Correy GJ, Gahbauer S, Fearon D, Wu T, Díaz RE, Young ID, Carvalho Martins L, Smith DH, Schulze-Gahmen U, Owens TW, Deshpande I, Merz GE, Thwin AC, Biel JT, Peters JK, Moritz M, Herrera N, Kratochvil HT, QCRG Structural Biology Consortium, Aimon A, Bennett JM, Brandao Neto J, Cohen AE, Dias A, Douangamath A, Dunnett L, Fedorov O, Ferla MP, Fuchs MR, Gorrie-Stone TJ, Holton JM, Johnson MG, Krojer T, Meigs G, Powell AJ, Rack JGM, Rangel VL, Russi S, Skyner RE, Smith CA, Soares AS, Wierman JL, Zhu K, O’Brien P, Jura N, Ashworth A, Irwin JJ, Thompson MC, Gestwicki JE, von Delft F, Shoichet BK, Fraser JS, Ahel I. (2021). Fragment binding to the Nsp3 macrodomain of SARS-CoV-2 identified through crystallographic screening and computational docking."},{"url":"https://doi.org/10.1126/sciadv.abo5083","text":"Correy GJ, Kneller DW, Phillips G, Pant S, Russi S, Cohen AE, Meigs G, Holton JM, Gahbauer S, Thompson MC, Ashworth A, Coates L, Kovalevsky A, Meilleur F, Fraser JS. (2022). The mechanisms of catalysis and ligand binding for the SARS-CoV-2 NSP3 macrodomain from neutron and x-ray diffraction at room temperature."},{"url":"https://doi.org/10.1016/s0076-6879(03)68015-8","text":"Snell EH, Bellamy HD, Borgstahl GE. (2003). Macromolecular Crystal Quality."},{"url":"https://doi.org/10.1002/prot.21246","text":"Meinhold L, Smith JC. (2007). Protein dynamics from X‐ray crystallography: Anisotropic, global motion in diffuse scattering patterns."},{"url":"https://doi.org/10.1038/s41467-023-36734-3","text":"Meisburger SP, Case DA, Ando N. (2023). Robust total X-ray scattering workflow to study correlated motion of proteins in crystals."},{"url":"https://doi.org/10.1126/science.1218231","text":"Karplus PA, Diederichs K. (2012). Linking Crystallographic Model and Data Quality."},{"url":"https://doi.org/10.1016/j.bpj.2026.08.008","text":"Nia SS, Saha A, Cho J, Zhou ZH, Ercius P, Mecklenburg M. (2026). Snapshots of internal protein crystal architecture at the nanoscale."},{"url":"https://onlinelibrary.wiley.com/doi/10.1002/pro.4224","text":"Winter G, Beilsten‐Edmands J, Devenish N, Gerstel M, Gildea RJ, McDonagh D, Pascal E, Waterman DG, Williams BH, Evans G. (2022). DIALS as a toolkit."},{"url":"https://doi.org/10.1145/3569951.3597559","text":"Boerner TJ, Deems S, Furlani TR, Knuth SL, Towns J. (2023). ACCESS: Advancing Innovation."}],"slug":"diffuse-scattering-east-west","social_posts_count":null,"social_posts_embed_src":null,"state":"PUBLISHED","subtitle":"Two teams measured three-dimensional diffuse scattering maps from SARS-CoV-2 Nsp3 macrodomain (Mac1) crystals to hone protocols and assess reproducibility. The resulting high-quality maps and new metrics to compare them will inform the interpretation of perturbation experiments.","tags":["Result"],"title":"Iterating diffuse scattering protocols in parallel leads to reproducible, high-quality data from Mac1","version_desc":"Fixed name: Advanced Light Source (ALS) beamline 8.3.1.","version_number":2,"versions":[{"id":525,"version_number":2,"version_desc":"Fixed name: Advanced Light Source (ALS) beamline 8.3.1.","doi":"10.82153/hrpy-t681.2","created_at":"2026-10-21T11:35:46.516Z"},{"id":516,"version_number":1,"version_desc":"v1 diffuse scattering East West","doi":"10.82153/hrpy-t681.1","created_at":"2026-10-20T19:47:36.821Z"}]}}