[Computer-aided prediction of the mutagenic activity substituted polycyclic compounds]

Area of Science:

• Computational chemistry
• Toxicology
• Structure-activity relationships

Background:

• Polycyclic compounds are prevalent in the environment and can exhibit mutagenic properties.
• Understanding the relationship between chemical structure and mutagenic activity is crucial for risk assessment.
• Predictive models can aid in identifying potentially hazardous compounds.

Purpose of the Study:

• To investigate the relationship between chemical structure and mutagenic activity in 54 polycyclic compounds.
• To compare the predictive performance of multiple linear regression (MLR) and artificial neural networks (ANNs).
• To identify key molecular descriptors influencing mutagenic activity.

Main Methods:

• Utilized multiple linear regression analysis and artificial neural networks for predictive modeling.
• Employed structural fragments, quantum chemical indices, and hydrophobicity (octanol-water partition coefficient) as molecular descriptors.
• Trained and validated models using a dataset of 54 polycyclic compounds.

Main Results:

• Both MLR and ANNs demonstrated accurate prediction of mutagenic activity for compounds similar to the training set.
• ANNs captured nonlinear relationships, potentially offering improved predictions for complex structures.
• The study substantiated the use of experimentally selected descriptors for mechanistic verification.

Conclusions:

• Computational approaches, including MLR and ANNs, are effective tools for predicting the mutagenic activity of polycyclic compounds.
• The choice of molecular descriptors significantly impacts model accuracy.
• Further research can leverage these findings to refine predictive toxicology and understand mutagenic mechanisms.