liveBook · Manning

Morgan fingerprint

As explained in Build AI Drug Discovery Pipelines

A Morgan fingerprint is a circular molecular fingerprint that encodes atom-centered structural neighborhoods within a specified radius.

A Morgan fingerprint is a circular molecular fingerprint that represents atom-centered neighborhoods within a chosen radius. It offers a practical molecular descriptor for screening, similarity searching, and machine-learning workflows. section 2.4.2 section 2.4.3

Generate the representation by:

  • Starting from every atom and expanding radially through neighboring atoms up to the specified radius. section 2.4.2
  • Hashing the resulting fragments into a fixed-length fingerprint, where binary bits record presence and count fingerprints record multiplicity. section 2.4.2
  • Choosing parameters explicitly: radius 2 and 2048 bits provide a local, computationally inexpensive default, while larger radii capture broader structure. section 2.4.3

RDKit featurization

RDKit can create binary and count-based fingerprints with GetMorganGenerator(radius=2, fpSize=2048). Applying this function across a molecule column produces descriptors for a compound library. section 2.4.2 section 2.4.3

Drug-repurposing dataset

A COVID Moonshot workflow computed 2048-bit, radius-2 Morgan fingerprints for 2,062 synthesized compounds before dimensionality-reduction analysis; optional IC50 labels separated compounds below or above 5 μM. section 7.1

Figure 2.13 Circular fingerprints look for patterns extending radially from each start position, rather than patterns of linear structures. The example shows how the search is conducted for a radius of 2 at three different start points.

Use the radial-neighborhood figure to see how circular fingerprints inspect structures around multiple atom-centered start points. section 2.4.2

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Explains neighborhood construction, hashing, parameter tradeoffs, multiplicity, collisions, and RDKit implementation, but not fingerprint reversal.
Examples
Uses RDKit code, radial fragment diagrams, fingerprint-bit visualizations, and COVID Moonshot compounds with IC50 measurements.
What it assumes
Assumes molecular structures available to RDKit, while defining atom neighborhoods, fingerprint bits, radii, and hash collisions during the treatment.
What you can do after
Apply it via RDKit to encode molecules as binary or count-based vectors for downstream analysis.
How it connects
Connects circular fingerprints with hashed fingerprints, extended-connectivity fingerprints, similarity searching, dimensionality reduction, and drug repurposing.

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.