Compound clustering

Group compounds by structural similarity, label the groups with NPClassifier natural-product classes, and explore them on an interactive map. With a screen, see which clusters are enriched for hits.

Library + screen hits The full compound library, plus the subset that were hits. One compound list Just cluster a single list of compounds.

Compound library

One SMILES per line (a header line and extra columns, such as names, are ignored). Up to 10,000 compounds.

…or paste here:

Screen hits

The hits as SMILES, one per line. They should be members of the library; any that are not will be added to it.

…or paste here:

Options

Distance cut

Tanimoto distance at which the tree is cut. Lower = tighter, more clusters.

Minimum cluster size

Smaller groups are reported as unclustered.

Label clusters with NPClassifier classes

Queries the public NPClassifier service for new structures, which can take a few minutes for large new libraries. Turn off for generic labels (“Cluster 1”, …).

Results are kept for 14 days and are available to anyone with the link.