Prediction of Biodegradability for Polycyclic Aromatic Hydrocarbons Using Various In Silico Modeling Methods

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Biotechnology

Background:

• Polycyclic aromatic hydrocarbons (PAHs) are persistent global environmental pollutants.
• Understanding the biodegradability of PAHs is crucial for environmental remediation.
• Predictive models can aid in assessing and managing PAH contamination.

Purpose of the Study:

• To establish quantitative structure-biodegradability relationships (QSBR) for PAHs.
• To develop in silico models for predicting PAH biodegradability.
• To identify key molecular descriptors influencing PAH biodegradation.

Main Methods:

• Utilized structural chemistry and quantum chemistry descriptors for molecular representation.
• Employed multiple linear regression (MLR), radial basis function neural network, and back-propagation artificial neural network (BPANN).
• Validated models using leave-one-out cross-validation.

Main Results:

• BPANN model achieved the highest correlation coefficient (R² = 0.9667) for the training set.
• Cross-validated correlation coefficients (q²) ranged from 0.6109 to 0.6887.
• The BPANN model demonstrated superior statistical significance compared to MLR and RBFNN.

Conclusions:

• QSBR models, particularly BPANN, can effectively predict PAH biodegradability.
• Molecular structure descriptors significantly influence the biodegradation of PAHs.
• These findings provide insights for environmental risk assessment and bioremediation strategies for PAHs.