Biomedical Engineering Reference
In-Depth Information
FIGURE 15.4 Biased similarity calculations. The influence of the molecular complexity
effect on fingerprint calculations is illustrated. Three molecules of increasing size are shown
(red, green, and blue, respectively). Each molecule is searched against the ZINC compound
database [104] and the resulting distributions of MACCS T c values are compared. As can
be seen, apparent similarity to database compounds increases with molecular complexity and
size. ( See insert for color representation of the figure. )
hence makes it more difficult to identify new hits. It has been shown that the use of
large and complex reference compounds generally leads to low hit rates [26], which
is a direct consequence of the complexity effect. In principle, the complexity effect
can be eliminated from similarity searching by using similarity metrics that weight
equally fingerprint bits that are set on and off [61] or by using fingerprints that yield
a constant bit density for all compounds, regardless of their topological complexity
and size [62]. The former approach might be controversial because in this case, the
absence of features contributes indirectly to similarity assessment.
Search WWH ::




Custom Search