PREDICTION OF SOIL SORPTION (KOC) OF HERBICIDES USING SEMIEMPIRICAL MOLECULAR-PROPERTIES

Relationships between soil sorption normalized to organic carbon (K(oc )) and molecular properties of 71 herbicides were examined. The K(oc) values were obtained from the literature. Various molecular properties were calculated by quantum mechanical methods using molecular modelin g software. The quantitative structure activity relationship (QSAR) mo dels based on four molecular properties, van der Waals volume (VDW(v)) , molecular polarizability (alpha), dipole moment (mu), and energy of highest occupied molecular orbital (e(HOMO)), together accounted for 7 0 % of the variation in K(oc). Herbiddes were broadly divided into six families based on structural similarities, and separate equations wer e established for each group. The three descriptors, VDW(v), alpha, an d mu, along with either energy of lowest unoccupied molecular orbital (e(LUMO)), or electrophilic superdelocalizability (S(E)), or e(HOMO) a ppeared to be determinants and accounted for 82 to 99 % of the variati on in K(oc). Applicability of these models was tested for one herbicid e analogue and 10 metabolites. The QSAR models appear to be specific t o structurally similar chemicals. The QSAR models could be developed t o predict K(oc) of structurally similar compounds even before they are synthesized or for some of the metabolites of existing herbicides. Mo dels of this type can also be developed to create priority lists for t esting, so that time, money, and efforts can be focused on the potenti ally most hazardous chemicals.