Proteoscope is an application for interactive 3D visualization of Protein Data Bank structures in legacy PDB (.pdb) and PDBx/mmCIF (.cif, .mmcif) formats. It is written as a single Go binary that embeds the browser UI, static assets, and bundled structure files, then serves the application on a local URL.

Proteoscope is designed for exploratory structural biology work: load a structure, inspect chains and ligands, switch molecular representations, color by scientific properties, view PDBx/mmCIF biological assemblies, search atoms or residues, measure distances, and export publication-prep screenshots.

Most users do not need Go installed. Download a prebuilt binary from the Proteoscope release page:

https://github.com/robert-mcdermott/proteoscope/releases/tag/v0.4

Choose the file for your operating system:

When Proteoscope starts, it prints a local URL, usually:

http://127.0.0.1:8765

It will also try to open that URL in your browser. For best performance, open the URL in Chrome, Edge, or Brave, which have strong WebGPU support. Safari may fall back to a slower compatibility renderer.

Download the macOS binary for your Mac:

Open Terminal and run:

chmod +x ./proteoscope-darwin-arm64
./proteoscope-darwin-arm64

Because the binary is not signed with an Apple developer certificate, macOS may show a warning such as:

Apple could not verify "proteoscope-darwin-arm64" is free of malware.

You can unblock it in either of these ways.

Option 1, from System Settings:

1. Try to open the app once and dismiss the warning.
2. Open System Settings.
3. Go to Privacy & Security.
4. Scroll to the Security section.
5. Click Open Anyway for Proteoscope.
6. Confirm by clicking Open.

Option 2, from Terminal:

cd ~/Downloads
xattr -d com.apple.quarantine ./proteoscope-darwin-arm64
chmod +x ./proteoscope-darwin-arm64
./proteoscope-darwin-arm64

If xattr says the quarantine attribute was not found, continue with the chmod and run commands. Replace the filename with the Intel binary if needed.

Download proteoscope-linux-amd64, then run:

cd ~/Downloads
chmod +x ./proteoscope-linux-amd64
./proteoscope-linux-amd64

If your browser does not open automatically, copy the printed localhost URL into Chrome, Edge, Brave, or another WebGPU-capable browser.

Download proteoscope-windows-amd64.exe, then double-click it.

If Windows SmartScreen blocks the app:

1. Click More info.
2. Click Run anyway.

If Windows marks the downloaded file as blocked:

1. Right-click proteoscope-windows-amd64.exe.
2. Choose Properties.
3. On the General tab, check Unblock if it appears.
4. Click Apply.
5. Run the .exe again.

You can also run it from PowerShell:

.\proteoscope-windows-amd64.exe

Proteoscope reads standard legacy .pdb coordinate files and modern PDBx/mmCIF files ending in .cif or .mmcif.

You can download more structures from the RCSB PDB download service:

https://www.rcsb.org/downloads

The RCSB download page supports downloading multiple files from the PDB archive and points users to individual data files from each structure's summary page. For modern RCSB downloads, PDBx/mmCIF (.cif or .mmcif) is usually the best choice. Legacy PDB coordinate files (.pdb) are also supported. Avoid PDFs, validation reports, sequence files, compressed archives, and documentation files unless you decompress or convert them into one of the supported coordinate formats first.

The dropdown labeled Bundled structure lists every supported structure file embedded from the repository's data/ directory at build time.

To add default structures to your own build:

1. Put one or more .pdb, .cif, or .mmcif files in data/.
2. Rebuild the Go binary.
3. The files will be embedded into the executable through Go's embedded filesystem.

Use Open local structure to load a structure from your computer.

The file is parsed in the browser. Proteoscope does not upload local coordinate files to an external server; the Go process only serves the local application.

The central viewport shows the active structure in 3D.

Mouse controls:

• Drag to rotate the molecule.
• Shift-drag to pan.
• Scroll to zoom.
• Click an atom to select it.
• Press R to reset the view.
• Press Esc to clear search and measurement state.

The renderer badge shows the active rendering mode:

• WebGPU: hardware-accelerated rendering is active.
• Canvas preview: WebGPU was unavailable in the current browser, so Proteoscope is using a compatibility renderer.

For best performance, use a browser with strong WebGPU support. Chrome, Edge, and Brave work well in current testing. Safari may fall back to a non-WebGPU path and can feel slower or less responsive with larger structures.

The top-left panel shows metadata and counts for the active structure.

The metadata strip reports:

• Coordinate format, such as PDB or PDBx/mmCIF.
• Experimental method when present.
• Resolution when present.
• Entry ID for deposited structures, or Local for files without an entry ID.
• Active assembly, either the asymmetric unit or a selected biological assembly.

For ordinary single-model structures, Atoms is the number of atoms in the loaded model.

For NMR ensembles or other multi-model coordinate files, Proteoscope renders one model at a time. In that case the UI shows:

• Atoms/model: atom count in the currently selected model.
• Total atoms: total coordinate records across all models.
• Models: number of models available in the file.

For example, an NMR ensemble with 40 models and 619 atoms in each model has 24,760 total coordinate records, but the viewer displays 619 atoms at a time.

Residues are grouped by chain, residue name, residue number, and insertion code. Standard residues, nucleic-acid residues, ligands, and waters are all recognized from the parsed coordinate records.

Chains come from the PDB chain identifier or, for PDBx/mmCIF, the author chain identifier when present. The chain panel lets you isolate one chain at a time or return to all chains.

Proteoscope provides four molecular representations.

Shows atoms as spheres and bonds as cylinders or ribbons. This is the best default for inspecting connectivity, ligands, cofactors, and residue-level geometry.

Shows atoms closer to their van der Waals size. This emphasizes molecular volume, steric packing, surface shape, and how tightly ligands or residues fill space.

Shows a simplified polymer trace. Proteins are traced through alpha carbons (CA), while nucleic acids are traced through backbone atoms such as P and sugar carbons. Ligands remain visible so cofactors and bound molecules are not lost.

Shows proteins as a smooth backbone cartoon generated from residue-level polymer chains. Alpha helices, beta strands, turns, and coils are distinguished from PDB HELIX/SHEET records or PDBx/mmCIF _struct_conf and _struct_sheet_range annotations when present. If no annotations are available, Proteoscope uses an approximate C-alpha geometry fallback and labels the source as computed. Cartoon mode keeps non-protein entities such as ligands, ions, waters, and nucleic acids available through the existing atom-level rendering controls.

Assigns a different color to each chain. This is useful for complexes, oligomers, protein-DNA assemblies, and comparing subunits.

Uses conventional element coloring, such as carbon, nitrogen, oxygen, sulfur, phosphorus, and common metal ions. This is useful for atom-level chemical inspection.

Colors atoms by B-factor, also called temperature factor or atomic displacement parameter. Lower and higher values are mapped to different colors so flexible or uncertain regions are easier to spot. For NMR files, B-factors may be zero or less meaningful depending on how the file was deposited.

Colors by residue class:

• Hydrophobic residues
• Polar residues
• Positively charged residues
• Negatively charged residues
• Nucleic-acid residues
• Ligands and solvent

This is useful for quickly reading chemical character across a structure.

Colors protein cartoons and protein atoms by secondary-structure assignment: alpha helices, beta strands, turns, and coils. Selection details show whether a residue's assignment came from file annotation, computed fallback geometry, or the default coil fallback.

Changes atom sphere size. Increase it for presentations or space-filling inspection; decrease it when dense structures become visually crowded.

Changes bond thickness. Increase it to make connectivity clearer; decrease it for very large structures.

Adds a subtle screen-space glow around atoms. This is a visual aid for depth and presentation screenshots. Set it lower for a quieter scientific plotting style.

Clips the structure along the current viewing direction. This helps inspect the interior of dense proteins, binding pockets, nucleic-acid grooves, or buried ligands.

Shows or hides HETATM records that are not water. This usually includes cofactors, ions, inhibitors, substrates, prosthetic groups, and crystallization components.

Shows or hides water molecules such as HOH, WAT, H2O, and DOD.

Water is off by default because crystallographic waters can clutter the view, but turning it on is useful for inspecting active sites, interfaces, and hydrogen-bonding networks.

Shows or hides hydrogen atoms. Hydrogens are often absent from X-ray structures and abundant in some NMR or modeled structures, so they are hidden by default.

Enables or disables gentle automatic rotation.

The search box finds residues, ligands, chains, atom names, elements, residue numbers, and atom serials.

Examples:

• heme
• HEM A142
• chain B
• lys 42
• FE
• CA

Search results include both residue-level and atom-level hits. Selecting a result focuses the view on that atom or residue representative.

Click an atom or choose a search result to select it.

The selection panel reports:

• Atom name
• Residue name and number
• Chain
• Element
• Atom serial number
• Occupancy
• B-factor
• Protein secondary-structure type and assignment source, when applicable

Use Frame to center the camera on the selected atom.

Proteoscope supports simple point-to-point distance measurements.

1. Click one atom.
2. Click a second atom.
3. Proteoscope records the distance between them in Angstroms (Å).

Each new atom click after the first creates a measurement from the previously selected atom to the newly selected atom. Measurement lines are drawn in the viewport and recent measurements appear in the selection panel.

Use Clear measure to remove measurement lines and reset the measurement sequence.

Common uses:

• Ligand-to-residue contact distances
• Metal coordination distances
• Hydrogen-bond candidate distances
• Interface contacts
• Nucleic-acid base-pair or intercalator geometry checks

Multi-model coordinate files, especially NMR ensembles, show a model slider at the bottom of the viewport.

Proteoscope displays one model at a time. Move the slider to inspect alternate conformations in the ensemble.

PDBx/mmCIF files may define biological assemblies in addition to the deposited asymmetric unit. When assembly definitions are present, Proteoscope shows a Biological assembly selector in the structure panel.

• Asymmetric unit shows the deposited coordinate set.
• Numbered assemblies apply the transformations from the mmCIF assembly records and display the generated biological unit.

Proteoscope reads _pdbx_struct_assembly, _pdbx_struct_assembly_gen, and _pdbx_struct_oper_list for assembly definitions. Operation expressions, including ranges and Cartesian-product expressions such as (1-4)(5,6), are resolved into 3D transforms before rendering.

Assembly generation can multiply atom counts dramatically, so Proteoscope uses a 300,000 atoms/model safety limit for generated assemblies. Assemblies above that limit are shown as unavailable rather than freezing the browser.

The chain panel lists all chains detected in the first model. Each row shows:

• Chain identifier
• Residue count
• Atom count
• Chain color

Click any chain row to toggle that chain on or off. You can show any arbitrary combination of chains, which is useful for comparing interfaces, hiding solvent or partner chains, or focusing on selected subunits. Press All to restore all chains.

Use the square toolbar button to export the current viewport as a PNG image. The image reflects the current camera angle, representation, coloring mode, visibility toggles, clipping, glow, and selection state.

Proteoscope supports two coordinate formats.

For .pdb files, Proteoscope reads these records and fields:

• HEADER, TITLE, EXPDTA, and resolution remarks for metadata.
• ATOM and HETATM for coordinates and atom properties.
• MODEL and ENDMDL for multi-model ensembles.
• HELIX and SHEET for secondary-structure hints.
• CONECT for explicit connectivity where present.
• Occupancy, B-factor, element, chain, residue name, residue number, insertion code, alternate location, and atom serial fields.

For .cif and .mmcif files, Proteoscope reads the PDBx/mmCIF categories used for interactive coordinate viewing:

• _atom_site for atom coordinates, element, atom name, residue name, chain, sequence ID, insertion code, occupancy, B-factor, model number, and ATOM/HETATM group.
• _entry, _struct, _struct_keywords, _exptl, _refine, _em_3d_reconstruction, and _reflns for title, entry ID, method, classification, keywords, and resolution metadata where present.
• _struct_conf and _struct_sheet_range for helix and sheet ranges.
• _struct_conn for explicit nonstandard links such as ligand, metal, disulfide, salt-bridge, or other curated structure connections when those records can be matched to rendered atoms.
• _pdbx_struct_assembly, _pdbx_struct_assembly_gen, and _pdbx_struct_oper_list for biological assembly generation.

For mmCIF chain display, Proteoscope prefers author-provided chain and residue identifiers (auth_*) when present because they usually match the identifiers researchers see in publications and RCSB pages. Label identifiers (label_*) are also retained internally for connection matching.

Bond handling:

• Explicit CONECT bonds are used when present.
• Explicit _struct_conn relationships are used for PDBx/mmCIF files when they can be resolved to atom records.
• Standard covalent bonds are inferred from element radii and interatomic distance because many coordinate files omit explicit records for ordinary polymer residues.

Alternate locations:

• Blank alternate locations are accepted.
• When alternate conformers describe the same atom, Proteoscope chooses the highest-occupancy conformer, then A/1, then the first available conformer.

Limitations:

• Biological assembly generation is currently implemented for PDBx/mmCIF files, not legacy PDB REMARK 350 records.
• Full solvent-accessible surfaces, electrostatics, density maps, and sequence annotation tracks are not yet implemented.
• PDBx/mmCIF coordinate, metadata, secondary-structure, and curated connection categories are supported; full dictionary coverage is intentionally out of scope for the interactive viewer.

• Go 1.24 or newer.
• A modern browser with WebGPU support. Chrome, Edge, and Brave are recommended for best performance. Safari may fall back to a compatibility renderer and is not recommended for large structures.

go run .

Useful flags:

go run . --no-open
go run . --host 127.0.0.1 --port 8765

go build -o proteoscope .

The resulting executable embeds:

• web/*
• supported structure files from data/, including .pdb, .cif, and .mmcif

That means the app can be distributed as one file.

GOOS=darwin GOARCH=arm64 go build -o dist/proteoscope-darwin-arm64 .
GOOS=darwin GOARCH=amd64 go build -o dist/proteoscope-darwin-amd64 .
GOOS=linux GOARCH=amd64 go build -o dist/proteoscope-linux-amd64 .
GOOS=windows GOARCH=amd64 go build -o dist/proteoscope-windows-amd64.exe .

go test ./...
node --test web/app.test.mjs

There is no Node build step. The frontend is plain embedded HTML, CSS, and JavaScript.

Your browser did not provide a WebGPU device. The app remains usable, but for the best rendering performance and shading quality, open the localhost URL in a browser with WebGPU enabled. Chrome, Edge, and Brave are the recommended choices. Safari may use a fallback path and can perform poorly on larger PDB files.

Proteoscope tries the requested port first and then searches nearby ports. Use the URL printed in the terminal.

Make sure the file is a coordinate file ending in .pdb, .cif, or .mmcif. Some RCSB download options provide compressed archives, validation reports, sequence files, or PDFs; those are not accepted directly by Proteoscope.

Proteoscope is licensed under the Apache License 2.0. See LICENSE for the full license text.