Application of Artificial Intelligence in Biodegradation Modelling

The inductive machine learning method coupled with a set of experts’ judgements and evaluated experimental biodegradation data were used to develop structural rules for ultimate biodegradation. Developed rules have been tested on BIODEG and MITI data. Both external validation tests have shown that the rules have solid predictive ability and performed better than other available methods. BIODEG and MITI data-bases were also used to develop new improved biodegradation rules. Two sets of developed biodegradation rules have very good classification ability, up to 90% for poorly biodegradable chemicals, and disclose structural features that either stimulate or hinder environmental biodegradation of organic chemicals.