Fig. 3

Cartoon of calculating UnBiological (Ub). This takes a ‘toy domain’ of four metabolites and three target molecules to explain the process. Only fragments of 3 or 4 atoms are considered in this example. In reality there are 611 metabolites, ~5000 targets molecules (note that a number of molecules are tested in more than one experimental series in Table 3) and 30912 Fragments of size 3 to 14 atoms. Metabolites and target molecules are used to generate Fragments that are present in at least two of the overall set of molecules (This is a convenient limitation on the number of Fragments, and may be revised in future implementations of the algorithm). Fragments are classified as to whether they occur in the set of metabolites (green) or do not occur in metabolites (red). The target set of molecules is then matched to the set of Fragments that do not occur in metabolites – the size of the largest such Fragment is the Ub measure. Note that in this simplified model it is clear that the presence of a chlorine atom confers ‘UnBiological-ness’ on a molecule. The size of the Ub fragment can be the same as the size of the whole molecule (e.g. 1-cloropropane in this example). As illustrated here, this approach takes no account of the potential reactivity of a molecule, only its topological structure