Uniqueness: skews bit occurrence frequencies in randomly generated fingerprint libraries

Gee Con Chen*

*Corresponding author for this work

Research output: Contribution to journalArticle


Requiring that randomly generated chemical fingerprint libraries have unique fingerprints such that no two fingerprints are identical causes a systematic skew in bit occurrence frequencies, the proportion at which specified bits are set. Observed frequencies (O) at which each bit is set within the resulting libraries systematically differ from frequencies at which bits are set at fingerprint generation (E). Observed frequencies systematically skew toward 0.5, with the effect being more pronounced as library size approaches the compound space, which is the total number of unique possible fingerprints given the number of bit positions each fingerprint contains. The effect is quantified for varying library sizes as a fraction of the overall compound space, and for changes in the specified frequency E. The cause and implications for this systematic skew are subsequently discussed. When generating random libraries of chemical fingerprints, the imposition of a uniqueness requirement should either be avoided or taken into account.

Original languageEnglish
Pages (from-to)741-745
Number of pages5
JournalMolecular Diversity
Issue number3
StatePublished - 1 Aug 2016


  • Algorithm
  • Fingerprints
  • Molecular modeling
  • Simulation

Fingerprint Dive into the research topics of 'Uniqueness: skews bit occurrence frequencies in randomly generated fingerprint libraries'. Together they form a unique fingerprint.

  • Cite this