Machine learning approaches for modeling the physiochemical characteristics of polycyclic aromatic hydrocarbons

Area of Science:

• Computational chemistry
• Cheminformatics
• Drug discovery

Background:

• Supervised machine learning (ML) methods are crucial for predicting bioactivity and structure-activity relationships in drug development.
• Quantitative structure-property relationships (QSPR) utilize molecular topological descriptors to model physicochemical properties.

Purpose of the Study:

• To identify key physicochemical properties impacting polycyclic aromatic hydrocarbons (PAHs).
• To establish algorithms linking topological indices to PAH physicochemical characteristics for enhanced prediction.

Main Methods:

• Application of supervised ML algorithms (Random Forests, Extreme Gradient Boosting).
• Utilizing eccentricity-based topological indices to represent molecular structures of PAHs.
• Developing QSPR models to correlate topological indices with physicochemical properties.

Main Results:

• Identification of significant physicochemical properties influencing PAHs.
• Successful construction of algorithms demonstrating the link between topological indices and physicochemical attributes.
• Validation of ML and QSPR combination for predicting molecular behavior.

Conclusions:

• ML and QSPR integration offers powerful computational tools for drug development.
• The developed models enhance understanding of PAH behavior.
• This approach supports future environmental forecasting and toxicological evaluations of PAHs.