On machine learning based QSPR analysis of amphetamine derivatives using regression models.

Area of Science:

• * Cheminformatics
• * Computational Chemistry
• * Medicinal Chemistry

Background:

• * Understanding the relationship between molecular structure and physicochemical properties is crucial for drug discovery.
• * Topological indices and polynomial regression offer potential for QSPR modeling.
• * Amphetamine derivatives are a class of compounds with significant pharmacological relevance.

Purpose of the Study:

• * To establish a quantitative structure-property relationship (QSPR) for amphetamine derivatives.
• * To evaluate the predictive power of neighborhood degree-based topological indices and NM-polynomials.
• * To compare polynomial regression models with Random Forest algorithms for property prediction.

Main Methods:

• * Calculation of neighborhood degree-based topological indices and NM-polynomials for amphetamine derivatives.
• * Development of polynomial regression models (cubic and quadratic) and Random Forest algorithms.
• * Prediction of physicochemical properties including boiling point, evaporation energy, flash point, molar refractivity, surface tension, polarizability, and SA (surface area).

Main Results:

• * Neighborhood-based indices effectively capture structural complexity, connectivity, and electronic characteristics relevant to stimulant behavior.
• * Cubic regression models demonstrated a better ability to represent nonlinear structural relationships compared to quadratic models.
• * Random Forest algorithms significantly improved prediction accuracy and generalizability, especially for properties dependent on molecular branching and electronic distribution.

Conclusions:

• * NM-polynomial based descriptors successfully correlate molecular topology with measurable physicochemical properties.
• * The developed QSPR models are valuable for computational property prediction, early drug screening, and cheminformatics-driven molecular design.
• * This approach offers a robust framework for understanding and predicting the behavior of stimulant-type molecules.