Ahmed W Eltayeb, Naeem Muhammad, Siddiqui Muhammad Kamran, Fiidow Mohamed Abubakar
Department of Mathematics and Statistics, College of Science, Imam Mohammad Ibn Saud Islamic University (IMSIU), Riyadh, Saudi Arabia.
Department of Mathematics, National University of Sciences and Technology (NUST), Islamabad, Pakistan.
Sci Rep. 2025 Aug 13;15(1):29667. doi: 10.1038/s41598-025-14134-5.
Dextran and chitosan, two natural polysaccharides, are recognized for their biocompatibility, biodegradability, and structural adaptability. Dextran, composed of glucose units with predominant α-(1→6) linkages, exhibits flexible conformations influenced by branching and molecular weight. Chitosan, derived from chitin via deacetylation, consists of β-(1→4)-linked D-glucosamine units and displays semi-crystalline behavior sensitive to pH and ionic conditions. An in-depth understanding of these structural properties is essential for applications in drug delivery, biomedical engineering, and polymer-based therapeutics. In this study, M-polynomial indices were calculated for dextran and chitosan using the edge/connectivity partition technique. Their predictive utility was evaluated through statistical correlations with several ADME-related physico-chemical properties of polycyclic drugs. Multiple regression models-Support Vector Regression, Lasso, Ridge, ElasticNet, and Multiple Linear Regression-were applied to model these relationships. Performance assessment was conducted using cross-validation and external test metrics, including the coefficient of determination ([Formula: see text]), Pearson correlation coefficient (R), root mean squared error, and p-values. Findings indicate that M-polynomial indices can reliably predict key properties such as molecular weight, exact mass, molar refractivity, polarization, complexity, and others. Several models demonstrated excellent predictive strength (e.g., [Formula: see text]) with statistical significance ([Formula: see text]), confirmed through both cross-validation and external validation. A Python-based tool was also developed to automate the computation of M-polynomial indices, enhancing efficiency and reproducibility. The results support the biological relevance of topological descriptors in modeling drug behavior and underline their potential utility in computational drug design, especially for biocompatible polysaccharide-based delivery systems.
葡聚糖和壳聚糖是两种天然多糖，因其生物相容性、生物降解性和结构适应性而受到认可。葡聚糖由具有主要α-(1→6)键的葡萄糖单元组成，其柔性构象受分支和分子量影响。壳聚糖是通过几丁质脱乙酰化得到的，由β-(1→4)连接的D-葡糖胺单元组成，并表现出对pH和离子条件敏感的半结晶行为。深入了解这些结构特性对于药物递送、生物医学工程和基于聚合物的治疗应用至关重要。在本研究中，使用边/连接性划分技术计算了葡聚糖和壳聚糖的M-多项式指数。通过与多环药物的几种与ADME相关的物理化学性质的统计相关性评估了它们的预测效用。应用了多种回归模型——支持向量回归、套索回归、岭回归、弹性网络回归和多元线性回归——来对这些关系进行建模。使用交叉验证和外部测试指标进行性能评估，包括决定系数（[公式：见原文]）、皮尔逊相关系数（R）、均方根误差和p值。研究结果表明，M-多项式指数可以可靠地预测诸如分子量、精确质量、摩尔折射率、极化率、复杂性等关键性质。几个模型表现出优异的预测强度（例如，[公式：见原文]），具有统计学意义（[公式：见原文]），通过交叉验证和外部验证得到证实。还开发了一个基于Python的工具来自动计算M-多项式指数，提高了效率和可重复性。结果支持了拓扑描述符在药物行为建模中的生物学相关性，并强调了它们在计算药物设计中的潜在效用，特别是对于基于生物相容性多糖的递送系统。
