> For the complete documentation index, see [llms.txt](https://guide.cryosparc.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://guide.cryosparc.com/processing-data/automated-workflows/full-automated-data-processing-a-case-study-using-gpcrs.md). # Full automated data processing: a case study using GPCRs ## Introduction Single particle cryo-EM is a valuable and growing technique for life science and drug discovery. Currently, obtaining state-of-the-art results from cryo-EM data analysis requires a human in the loop to analyze intermediate results and make image processing decisions. This bottleneck limits the achievable throughput of structure determination, especially in high-throughput settings such as structure-based drug design. In this work, we develop an end-to-end automation strategy for repeat-target structure determination using new tools in CryoSPARC. {% hint style="success" icon="link" %} Read the entire preprint here: [End-to-end automation of repeat-target cryo-EM structure determination in CryoSPARC](https://doi.org/10.1101/2025.10.17.682689) {% endhint %} We demonstrate completely hands-off processing of 21 challenging G protein-coupled receptor (GPCR) datasets. In 17 of 21 cases, automated processing meets or exceeds published resolution and map quality and, in several cases, provides significant improvement in receptor and ligand density that allows improved model building. Our results on both active and inactive state GPCRs show that our automation strategy generalizes easily to new target classes, and that complete automation of data processing is straightforward to achieve in CryoSPARC. We provide downloadable CryoSPARC Workflow files so that users can import, replicate, adapt and extend our automated workflow for their own targets, enabling cryo-EM to be applied at larger scales and to answer larger biological questions. ## GPCRs: A challenging test case for automation GPCRs are small, real-world drug targets and are generally challenging for single particle analysis. We selected 21 EMPIAR datasets of GPCR complexes in both the active and inactive state (see [Table 1 in the preprint](https://doi.org/10.1101/2025.10.17.682689) for more details). ## Development of a single-particle automation strategy: Workflow v1 We detail the development of the generalized automation strategy in the [preprint](https://doi.org/10.1101/2025.10.17.682689). The strategy we ultimately produced (Workflow v1) is illustrated as a flowchart in the figure below. Each stage of the workflow and each new method or tool that was developed in CryoSPARC to support automated data processing, are described in the Methods section of the preprint. {% hint style="success" icon="link" %} Read the entire preprint here: [End-to-end automation of repeat-target cryo-EM structure determination in CryoSPARC](https://doi.org/10.1101/2025.10.17.682689) {% endhint %}

Workflow v1. Complete workflow for end-to-end automation of repeat-target structure determination. A. Workflow flowchart from import to final refinements, detailing required inputs at each stage. Dataset level inputs must be provided for each new dataset. Class level inputs (C) only need to be set once, and are reused for all datasets in the target class. For example, the low resolution reference map and corresponding mask for each target class in this work are shown in insets. Workflow level inputs (W) can be reused across multiple classes. B. 2D classes for template picking are automatically selected using Reference Based Auto Select 2D (RBAS 2D). C. Micrograph Denoiser improves particle picking, and particles close to junk and contaminants are automatically rejected by Micrograph Junk Detector. D. Decoy classification curates particles in 3D without introducing orientation bias. E. Further curation using Ab-initio Reconstruction and Heterogeneous refinement yields final high-quality particles. F. Non-Uniform Refinement (left) produces an optimal final global refinement, and Local Refinement (right) using the input mask improves map quality in the receptor region.

## Results of automated processing on GPCRs The [results](https://doi.org/10.1101/2025.10.17.682689) demonstrated that we were able to successfully produce maps with resolution and map quality equal or better than manual processing in the repeat-target scenario, with no manual intervention. The automation strategy worked successfully even in the context of datasets containing poor quality micrographs, contaminants, aberrations, aggregation, denatured and junk particles. In multiple cases, the automated workflow yielded improved map quality and interpretability in the receptor region where a ligand is bound, demonstrating that the strategy is realistic for real-world drug discovery scenarios. Finally, the automated workflow could be completed in a day or less on a 2-GPU workstation, with further speedups possible by adding parallel compute resources.

Automated processing matches or exceeds the map resolution and map isotropy compared to deposited manual processing results for majority of GPCR datasets. Left: Gold-standard Fourier Shell Correlation (FSC) and right: Conical FSC Area Ratio (cFAR). Values from deposited published results shown as gray circles. Values from automated processing shown as filled, colored circles: improved (green), equivalent (blue), or worse (orange). Our automated workflow achieves equal or better resolution in 17 out of the 21 datasets with six improvements of 0.3Å or greater. FSC and cFAR were measured using a single soft mask for each dataset, so that the values are directly comparable. Asterisks indicate datasets for which half-maps were not deposited to EMDB, so the deposition-reported resolution is used instead and cFAR scores could not be calculated.

An excerpt of results from this work showing the deposited maps and our automated processing maps at similar thresholds for each dataset. For each global map, FSC resolutions (upper left) and cFAR score (lower right) are shown. For automated maps, color scheme denotes improved (green), equivalent (blue), or worse (orange) relative to the manually processed deposited maps. Ligand density is shown with the published atomic model docked for deposited maps, and a re-built atomic model docked for our automated results, to emphasize the improved map quality and interpretability. EMPIAR-10673 demonstrates this workflow can cleanly recapitulate a nearly 2Å result. The ligand density for EMPIAR-10855 is more fully resolved, for EMPIAR-11119, we are actually able to recover density for prostaglandin E2 where the depositors could not, but suspected was there. Lastly, for EMPIAR-11433, where automated processing resolution is nominally lower than that of the depositors by 0.2Å, we still recovered receptor and ligand density that is nearly indistinguishable from the depositors.

{% hint style="success" icon="link" %} Review all the results here: [End-to-end automation of repeat-target cryo-EM structure determination in CryoSPARC](https://doi.org/10.1101/2025.10.17.682689) {% endhint %} ## Using the GPCR workflow file on your own GPCR datasets The instructions below outline how to take our GPCR Workflow file and use it on your own GPCR datasets (i.e., targets of the same “class”). 1. Download the workflow JSON files, volumes, and masks used in the whitepaper: [GPCR\_automated\_Workflow\_materials\_v1.zip](https://structura-assets.s3.us-east-1.amazonaws.com/automated-workflows/GPCR_automated_workflow_materials_v1.zip) 2. [Upload these volumes](https://guide.cryosparc.com/processing-data/automated-workflows/practical-tips-uploading-files-and-using-workflows#uploading-volumes-and-masks-for-use-in-workflow) to your compute setup. They do not need to be located within the project. 3. [Import workflow JSON file](https://guide.cryosparc.com/processing-data/automated-workflows/practical-tips-uploading-files-and-using-workflows#importing-the-workflow-json-file) into CryoSPARC using the workflow panel in the sidebar. 4. [Select the workflow from the list](https://guide.cryosparc.com/processing-data/automated-workflows/practical-tips-uploading-files-and-using-workflows#applying-the-workflow), set all parameters, and choose the compute node to utilize. Parameters that might need to be adjusted (on a per dataset basis) are below: Import movies * Movies data path * Gain reference path * Raw pixel size (Å) * Accelerating voltage (kV) * Spherical aberration (mm) * Total exposure dose (e/Å^2) * Flip gain ref and defect file in Y * Exposure group metadata (if present) Import 3D Volumes * Path to volumes/masks for GPCR reference, junk volumes, and the receptor mask Patch Motion Correction * Output F-crop factor Exposure Group Utilities * Use a regular expression string * Change this regular expression string to one of the following: * If testing on an EMPIAR dataset in used in this study, refer to Table S1 for the correct regular expression to use. * If using another dataset that captures the beam image-shift group in the filename, use a regular expression that would separate them. * If beam image-shift metadata was imported with the movies, then set this job to the cluster and split method and supply the number of expected groups and other parameters that might be needed. * If beam image-shift was not used during data collection, the regular expression parameter can be left blank. Extract from micrographs (x2) * Extraction box size Note: if using our reference, this should be 310.6 divided by the pixel size of the motion corrected micrographs (rounded to the nearest even pixel - e.g. 361.18 —> 362). 5. Click on the green “Apply” button in the bottom of the workflow GUI. 6. Download and inspect your final volumes, and proceed with any advanced processing if desired.