tanimoto

KNIME Analytics Platform
1

As far as I understand the Jaccard/tanimoto coefficient measures the interaction/union of fingerprint bits and thus is a similarity measure. Why then is 1-tanimoto used as a similarity measure in the "Similarity Search" node, since tanimoto would just be a distance measure then?

Can someone clarify, since using it as a cutoff would affect the outcome of chemical similarity searches? See en.wikipedia.org/wiki/Jaccard_index

 

 

2

The distance function you select at the top of the dialog is always a distance function. Also "Tanimoto" refers to the Tanimoto distance and not the coefficient. Therefore "1 - tanimoto distance" is a similarity.