This component groups a list of unique compounds into molecular series according to the series identification method selected in the Molecular Series Configuration component. It accepts as input a list of unique compounds with their molecular structure in SMILES or SDF format and, optionally, an associated biological activity. Upon grouping, activities are aggregated using multiple methods.
The component assumes a unique list of molecules as input (which can be obtained using the Molecule Activity Aggregation component), with the molecule ID as the row ID.
Workflow position
This component must be preceded by the Molecular Series Configuration component and should be followed by the Molecular Series Tuning component for manual adjustment of the series definition (in the case of clustering). The resulting series can then be analyzed and selected using the Molecular Series Selection component.
Clustering
When the selected series type involves clustering, a hierarchical clustering is performed using:
- Distance metric — Cosine distance computed from Morgan count-based radius-2 1024-bit hashed molecular fingerprints
- Linkage strategy — Complete Linkage
The optimal number of clusters is determined by default using the Silhouette method, but can be easily adjusted using the Molecular Series Tuning component. The distance matrix and the resulting clustering model are provided as the 2nd and 3rd outputs, respectively.
The component assumes a unique list of molecules as input (which can be obtained using the Molecule Activity Aggregation component), with the molecule ID as the row ID.
Workflow position
This component must be preceded by the Molecular Series Configuration component and should be followed by the Molecular Series Tuning component for manual adjustment of the series definition (in the case of clustering). The resulting series can then be analyzed and selected using the Molecular Series Selection component.
Clustering
When the selected series type involves clustering, a hierarchical clustering is performed using:
- Distance metric — Cosine distance computed from Morgan count-based radius-2 1024-bit hashed molecular fingerprints
- Linkage strategy — Complete Linkage
The optimal number of clusters is determined by default using the Silhouette method, but can be easily adjusted using the Molecular Series Tuning component. The distance matrix and the resulting clustering model are provided as the 2nd and 3rd outputs, respectively.