Glossary

Molecular fingerprint

A fixed-length binary or count vector that encodes which structural features are present in a molecule — the standard way to convert chemical structures into numerical inputs for machine learning models.


A molecular fingerprint converts a chemical structure into a numerical vector that machine learning models can process. Each bit (or count) in the vector represents whether a particular structural feature — a specific substructure, atom environment, or ring pattern — is present in the molecule. Two molecules with similar fingerprints tend to have similar properties, which is the underlying assumption that makes fingerprint-based ML useful.

Common types:

  • ECFP (Extended Connectivity Fingerprints): the most widely used family; encode circular neighborhoods around each atom out to a specified radius. ECFP4 and ECFP6 are standard choices for property prediction.
  • MACCS keys: a fixed 166-bit fingerprint based on predefined structural keys; simpler and interpretable but less expressive than ECFP.
  • RDKit fingerprints: open-source implementation with several variants; default choice for researchers using Python.

How they’re used in AI chemistry:

Fingerprints are the typical feature representation for molecular property prediction models (Chemprop, random forests, support vector machines). They’re also used to measure molecular similarity — a Tanimoto coefficient between two fingerprints gives a standardized 0–1 similarity score used in virtual screening and dataset analysis.

Limitations:

  • Fingerprints encode topology but not 3D shape — two molecules can have identical fingerprints but different conformations and binding behaviors
  • Fixed-length vectors discard some structural information; graph neural networks (like those in Chemprop) learn their own representations from molecular graphs and often outperform fingerprints on complex property prediction tasks
  • Similarity in fingerprint space doesn’t always translate to similarity in biological activity (the activity cliff problem)

Related terms: SMILES Notation, Graph Neural Network

Related guide: Chemistry