The Cluster Ligands protocol in Discovery Studio lets you quickly organize a set of input ligands into meaningful groups (clusters) based on one or more molecular descriptors. You can choose numeric properties (e.g., AlogP, molecular weight, H-bond donors/acceptors), fingerprint-based similarity (e.g., FCFP_6, MDL public keys), or a combination of both. The protocol also accepts properties defined by the user or taken from the molecular properties table.

Ligands can be clustered by:

• Fixed Number of Clusters: Number of cluster centers.
• Fixed Number of Molecules per Cluster: Average number of molecules per cluster.

Ligand clustering can be very useful in recognizing the common pattern shown by active compounds in a high throughput screening. Other usages include choosing a small subset to represent a large data set. 

What you get back:

• The input ligands, each labeled with the cluster it belongs to

• An sd file with the center of each cluster

Where it helps:

• Making sense of HTS hits: Filter the labeled ligands by cluster to see which chemotypes recur among the actives. Several hits in one cluster are stronger evidence of real SAR than singletons, and singletons can be checked first as possible artifacts.

• Representing a large library with a small set: The sd file with cluster centers is a ready-made diverse subset. Use it to choose compounds for follow-up assays, purchasing or a pilot screen, instead of sampling at random.

• Combining structure and properties: Cluster on a fingerprint plus properties such as AlogP, so groups reflect both scaffold similarity and physicochemical profile.