Graph-Theoretical Approach for Predicting Physicochemical Properties of Stiff-Person Syndrome Drugs

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Neurology

Background:

• Stiff person syndrome (SPS) is a rare neurological disorder characterized by muscle rigidity and spasms.
• Understanding the molecular underpinnings of SPS is crucial for developing effective treatments.

Purpose of the Study:

• To apply quantitative structure-property relationship (QSPR) analysis to identify molecular descriptors associated with SPS.
• To explore the utility of graph-theoretical methods and machine learning in drug discovery for SPS.

Main Methods:

• Calculation of topological indices (e.g., Zagreb indices) and M-polynomials.
• Application of various regression models (linear, quadratic, exponential, power) and heatmap correlation analysis.
• Utilizing gradient boosting, an ensemble machine learning method, for predictive accuracy and descriptor importance evaluation.

Main Results:

• Identified significant correlations between topological indices and physicochemical properties relevant to SPS.
• Gradient boosting enhanced predictive accuracy and highlighted key molecular descriptors.
• M-polynomials and their graphs provided insights into structural features of potential therapeutic agents.

Conclusions:

• Graph-theoretical methods are valuable tools for understanding SPS molecular characteristics.
• Computational approaches, including QSPR and machine learning, can advance the design of targeted SPS therapies.
• This study contributes to the drug discovery pipeline for rare neurological conditions like SPS.