Moving beyond the van Krevelen Diagram: A New Stoichiometric Approach for Compound Classification in Organisms

van Krevelen diagrams (O/C vs H/C ratios of elemental formulas) have been widely used in studies to obtain an estimation of the main compound categories present in environmental samples. However, the limits defining a specific compound category based solely on O/C and H/C ratios of elemental formulas have never been accurately listed or proposed to classify metabolites in biological samples. Furthermore, while O/C vs H/C ratios of elemental formulas can provide an overview of the compound categories, such classification is inefficient because of the large overlap among different compound categories along both axes. We propose a more accurate compound classification for biological samples analyzed by high-resolution mass spectrometry based on an assessment of the C/H/O/N/P stoichiometric ratios of over 130 000 elemental formulas of compounds classified in 6 main categories: lipids, peptides, amino sugars, carbohydrates, nucleotides, and phytochemical compounds (oxy-aromatic compounds). Our multidimensional stoichiometric compound classification (MSCC) constraints showed a highly accurate categorization of elemental formulas to the main compound categories in biological samples with over 98% of accuracy representing a substantial improvement over any classification based on the classic van Krevelen diagram. This method represents a signficant step forward in environmental research, especially ecological stoichiometry and eco-metabolomics studies, by providing a novel and robust tool to improve our understanding of the ecosystem structure and function through the chemical characterization of biological samples.范克雷维伦图（van Krevelen diagrams，即以元素式的氧碳比与氢碳比为坐标轴）已被广泛应用于环境样品中主要化合物类别的估算研究。然而，此前从未有研究基于元素式的氧碳比与氢碳比，精准界定各类化合物类别的划分阈值，以用于生物样品中的代谢物分类。此外，尽管以元素式的氧碳比与氢碳比可概览化合物类别，但由于不同化合物类别在两个坐标轴上均存在大量重叠，此类分类效率低下。我们针对经高分辨质谱（high-resolution mass spectrometry）分析的生物样品，提出了一种更精准的化合物分类方法：该方法基于对13万余种已归入六大主要类别的化合物的元素式碳、氢、氧、氮、磷化学计量比的评估，这六大类别分别为脂质、肽类、氨基糖、碳水化合物、核苷酸以及植物化学物质（含氧芳香族化合物）。我们提出的多维化学计量化合物分类（multidimensional stoichiometric compound classification, MSCC）约束条件，可将生物样品中的元素式精准归类至对应主要化合物类别，分类准确率超过98%，相较于基于经典范克雷维伦图的分类方法实现了大幅提升。该方法为环境研究领域，尤其是生态化学计量学与生态代谢组学研究，提供了一种新颖且稳健的工具，可通过生物样品的化学表征加深我们对生态系统结构与功能的理解，堪称环境研究领域的一大重要进展。