Abstract
The toxicity to Chlorella vulgaris, expressed as log(1/EC50), of two sets of aromatic compounds has been examined. For the first set, which consists of 13 mono- and di-substituted nitrobenzenes, it was found that one- or two-descriptor models provide useful correlations. A six-descriptor regression model for the log(1/EC50) values of a larger set that consists of 49 aromatic compounds has also been developed. Overall, it was found that a combination of a small number of trivial 'feature count' classical descriptors (numbers of atoms of a particular type) and less traditional quantities (entropy-like momentum-space descriptors) have potential benefits for useful QSAR models.