liveBook · Manning

Hierarchical clustering

As explained in Build AI Drug Discovery Pipelines

Hierarchical clustering is a clustering technique that builds a hierarchy of nested groups by iteratively merging or splitting clusters.

Hierarchical clustering organizes compounds into nested groups by repeatedly merging similar clusters or splitting larger clusters. Agglomerative clustering begins with one cluster per compound and merges the closest clusters until one hierarchy remains; divisive clustering starts with all compounds together and splits them. For molecular libraries, representations, distance metrics, clustering choices, and sampling goals must be considered together. A practical default uses Morgan fingerprints with Tanimoto distance (1 minus similarity), while linkage choices include single, complete, average, and Ward’s method. section 7.2.3

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Explains agglomerative and divisive procedures, linkage choices, fingerprint implementation, dendrograms, and sampling, but not clustering theory in depth.
Examples
Uses RDKit fingerprints, scikit-learn models, dendrograms, representative sampling, and diversity-versus-redundancy evaluation.
What it assumes
Assumes molecules can be represented as fingerprints and compared by similarity, while defining clustering terminology before implementation.
What you can do after
Apply it via RDKit and scikit-learn to cluster compounds and sample representatives for library screening.
How it connects
Connects molecular fingerprints, Tanimoto similarity, linkage methods, dendrograms, cluster sampling, and Taylor-Butina clustering.

Also explained in

Compare how 8 books explain hierarchical clustering →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.