The hypothetical modal object, for binary data, is the hypothetical object that posseses the commonest state for each characteristic.
Sokal, RR; Sneath PHA (1963) Principles of Numerical Taxonomy W.H.Freeman &Co San Francisco p195
In the Daylight world a modal fingerprint is created for a set of compounds by setting a bit in an empty fingerprint where the corresponding bit is set in the majority of the fingerprints of the compounds in the set. The resultant fingerprint is a valid target for comparison in an estimate of similarity, but may not be the fingerprint of a valid molecule. In some cases it may represent the 'parent' structure of a set of congeneric compounds, or the common, possibly disconnected pharmacophore of a drug set, but this cannot be relied on.
For example if you take the 41 compounds from medchem03
with morphine in the name, the modal fingerprint has 0.9980 Tanimoto
similarity to the fingerprint of morphine. No other member of the set has
higher similarity to the modal.
If medchem03 is clustered, using k-modes, similar behaviour is observed. However in other databases with less alkaloids, morphine ends up on the edge of clusters dominated by polycyclic phenolics.
The use of the hypothetical modal object as the nearest to the ideal that could be obtained, is described in Liston, J; Weibe, W; Colwell, RR; (1963) J. Bacteriol 85, 1061-1070
Daylight Chemical Information Systems, Inc.