User manual

CBCSpace projects chemical libraries into 2D or 3D maps, so you can see how compounds cluster, where two libraries overlap, and which molecules sit in a given region. Everything runs on the server you are connected to; uploaded files stay with your browser session.

1. Quick start

1. The left panel lists the default libraries. Tick the ones you want to compare.
2. Choose a feature space (descriptors or fingerprints) and a projection method.
3. Press ▶ compute projection. The plot appears on the right.
4. Hover over points to see structures. Use the search box at the top of the plot to find a compound by name.

2. Datasets

Default libraries

Default libraries are shared by everyone on this server. They are grouped into folders in the sidebar. Click a folder name to open or close it. The tick box beside a folder shows or hides all of its libraries at once; a dash means only some are shown. The all · none links above the list show or hide every default library.

Your own libraries

Drop a file onto the drop csv / tsv / excel area, or click it to pick a file. Your library is added under your libraries. It is only visible to you, in this browser session. Use ✕ next to a library to remove it.

Show, colour and compare

• The tick box next to a library shows or hides it on the plot.
• The coloured dot opens a colour list. Each library keeps its colour in the plot legend.
• Libraries that are shown are projected together, so overlap and spread can be compared directly.

3. Feature space

Structures are cleaned when they are loaded: salts and solvents are removed and charges are neutralised, so counter-ions do not change the map. Molecules that cannot be parsed, or that have a missing descriptor, are left out.

4. Projection

• 2D / 3D switches between a flat map and a rotatable 3D map. Box and lasso selection work in 2D only.
• For UMAP and t-SNE on fingerprints, PCA pre-reduce to 50D makes the computation much faster. It uses Euclidean distance on the reduced data instead of Tanimoto on the raw bits.
• The status message in the header reports the method, dimensions, fingerprint type and number of molecules.

5. Lipinski filters

After a projection, sliders for MW, cLogP, HBD and HBA appear. Moving them hides molecules outside the range from the plot; the counts are updated. Ro5 defaults sets the upper limits to the usual rule-of-five values (MW 500, cLogP 5, HBD 5, HBA 10). reset restores the full range.

6. Reading the plot

• Hover over a point to see its structure, name, dataset and coordinates in a card. Move away to close it.
• Click a selected molecule in the selection list to pin its card in place.
• Zoom and pan with the mouse wheel and by dragging (2D: pan mode; 3D: rotate). ⤢ reset view returns to the full map.
• The point size slider changes the size of all points.
• When points are hidden by sampling or filters, a note in the top-right corner of the plot says how many are shown.

7. Searching

The search box at the top of the plot finds molecules by identifier (name). It searches every shown library and lists up to 50 matches. Click a match to highlight it on the plot and zoom to it. If the molecule was not included in the displayed sample, the card still shows, with a message saying it is not on the plot. Press Esc or ✕ to clear the search and the highlight.

8. Selecting and exporting

• In 2D, choose ▭ (box select) or ➰ (lasso), then drag on the plot. Pan mode (✥) turns selection off.
• The selection panel at the bottom right lists the selected molecules (first 500 shown) with their library.
• ⬇ export downloads the selection as cbcspace-selection.csv: dataset, identifier, SMILES and the plotted coordinates. ✕ clears the selection.

9. Publishing a default library

If the server has publishing enabled, a ★ button appears next to your own libraries. It makes the library a shared default for everyone.

1. Click ★. Enter a name for the library.
2. Enter the admin token. Your browser remembers it after a successful publish.
3. The library is added to the Other folder. Ask the administrator to move it into a folder if needed.

10. Sampling and limits

• Coordinates are computed for every molecule, but at most 150,000 points are drawn. When the library is larger, the plot is sampled, and the note says so. Overlap and spread figures use all molecules.
• UMAP and t-SNE are fitted on a random sample of up to 50,000 molecules; the remaining molecules are placed onto that map. Results can therefore differ slightly between runs on very large libraries.
• Large projections can take a while. The progress bar shows when a run is in progress.

11. Input files

• Accepted formats: .csv, .tsv, .xlsx, .xls. Maximum size 2 GB.
• One column must hold SMILES. Its header is usually smiles, smi, canonical_smiles or structure; otherwise the first column that reads as SMILES is used.
• An optional name or ID column (for example name, id, compound_id or chembl_id) gives each molecule its label. Without one, molecules are numbered as <file>_1, <file>_2, and so on.
• Rows with invalid SMILES are dropped, and the upload summary shows how many molecules were kept.

12. Troubleshooting

The page you are reading is the manual. The explorer itself is at the main page.