CryoCore gives your agent the tool knowledge and workflow instructions to plan and run computational experiments in cryo-EM and cryo-ET. Point your agent at this repository with your scientific question, data, and compute limits. Ask it to reconstruct a protein complex, compare candidate conformations, or test which analysis method works best on your data.
Give your agent the repository link, your question, and access to your data.
I have cryo-EM movies and sequences for a protein complex. Use
https://github.com/BioSymphony/cryocore to reconstruct a map and build
a model. Choose the tools, set them up, and run the analysis.
Show me how well the model fits the density, flag regions that need
review, and save the commands so I can reproduce the results.
The CryoCore skill gives your agent the workflow instructions. Include your available compute and budget with the request.
The software registry covers over 100 tools and resources. Records track tool roles, upstream sources and versions, setup and hardware requirements, license terms, and readiness. Your agent uses these records to select tools and prepare the required runtimes.
| Task | Tools in the catalog |
|---|---|
| Find public data | EMPIAR for image sets, EMDB for maps, RCSB PDB for coordinates, and the CryoET Data Portal for tomograms and annotations. Accession APIs |
| Correct movies and check micrographs | MotionCor3 for motion correction, GCtfFind or CTFFIND for contrast transfer function (CTF) estimation, and Miffi for image quality scores. Micrograph QC |
| Pick particles and reconstruct a map | Topaz and crYOLO for picking, RELION and CryoSPARC for classification and refinement, and pyem for converting CryoSPARC metadata to STAR. Processing contract |
| Process tilt series | AreTomo3, WarpTools, and IMOD for alignment and tomogram reconstruction. |
| Locate features in tomograms | PyTom Match Pick for template matching, OCTOPI for particle proposals, ETSAM for membrane segmentation, and copick for managing picks and annotations. |
| Postprocess maps | LocScale and CryoCodex for map enhancement, and IsoNet2 for tomogram restoration. |
| Build and refine models | ModelAngelo, CryoAtom, and EM3DFold for building atomic models, Coot for inspecting and editing them, and Servalcat and Phenix for refinement. Model-building contract |
| Check map support | EMDA for map comparison, DAQplugin and checkMySequence for local model and sequence-register checks, and wwPDB for deposited validation reports. Map/model review |
| Compare conformations | cryoDRGN, RECOVAR, and DynaMight for single-particle heterogeneity, and tomoDRGN for subtomogram heterogeneity. Heterogeneity review |
| Make figures and review scenes | PyMOL and ChimeraX for map/model overlays, Mol* for browser views, Blender for animation, and matplotlib for Fourier shell correlation (FSC) plots. Rendering guide |
Your agent can assemble these workflows using CryoCore's scripts, workflow contracts, and upstream tools.
Start with acquired movies, microscope metadata, and the specimen sequence. Run MotionCor3 for alignment, GCtfFind for CTF estimation, and Topaz for picking. RELION combines corrected micrographs, CTF metadata, and pick coordinates for extraction, classification, and refinement. Use the map and sequence with ModelAngelo to build a candidate model, then refine it against the matching half-maps with Servalcat and render overlays in ChimeraX.
Request particle counts at each stage, maps and half-maps, model coordinates, fit and geometry reports, and reproducible figures. Preserve particle IDs and half-set membership through processing.
Reconstruct acquired tilt series with AreTomo3. Supply a density template and mask to PyTom Match Pick to locate candidate particles. In parallel, evaluate ETSAM membrane segmentation. Register the tomogram, picks, and masks in copick with matching voxel sizes and coordinate frames.
Request scored particle coordinates, membrane masks, and distances between candidate particles and membranes. Use independent annotations to evaluate localization and segmentation.
Give cryoDRGN and RECOVAR the same particle cohort with compatible orientation and CTF metadata, plus the mask required by the selected method. Select candidate maps by particle membership or a specified structural feature. Use EMDA to compare the maps at a common grid and resolution, then render the differences in ChimeraX.
Request map comparisons, associated particle sets, and a repeat analysis on independent subsets. Identify which differences persist across methods and repeated runs.
For EMDB map EMD-43816 and PDB model 9ASJ, retrieve the map, coordinates, and wwPDB validation report. Check map/model alignment and density support, then render weakly supported regions for inspection. Return a review with source accessions, local fit metrics, and annotated figures.
Test whether changing the particle picker improves the reconstruction.
Would Topaz or crYOLO give us a better reconstruction from these movies?
Use CryoCore to plan a fair comparison and run it. Keep some annotated
micrographs out of training so we can check the picks independently.
Refine both particle sets in RELION with matched particle counts and
settings, then compare the maps and half-map FSC.
Estimate the compute and storage we'll need for the full dataset.
Run within the budget I've given you, and show me where the methods
agree, where they differ, and whether the differences hold up.
Use the same approach to compare model builders against an experimental map, evaluate map enhancement against original half-maps, or plan a parameter sweep on a Slurm cluster.
Each handoff carries the data and metadata the next tool needs. For motion correction followed by CTF estimation, the corrected pixel size must follow the micrograph.
Your agent runs bundled scripts through its terminal and invokes upstream tools in the environments it prepares. CryoCore includes adapters for micrograph scoring and motion/CTF comparisons, accession helpers, renderer scripts, and output checks. Callable scripts
Use the tool-use guide for commands and format conversions. It explains how to match units and coordinate frames before passing outputs to the next tool. Tool use and chaining
- Agent quickstart for task-specific skills.
- Workflow modules and recipes for planning experiments.
- Compute backends for local, cloud, or HPC setup.
- Toolwatch for recent releases and preprints.
- Documentation and MIT license.
