Zhang Meng, Li Xiang, Oladeinde Adelumola, Rothrock Michael, Pokoo-Aikins Anthony, Zock Gregory
Department of Mathematics, University of North Georgia, 82 College Cir, Dahlonega, GA 30597, USA.
U.S. National Poultry Research Center, Egg & Poultry Production Safety Research Unit, Agricultural Research Service, U.S. Department of Agriculture, 950 College Station Road, Athens, GA 30605, USA.
Microorganisms. 2024 Sep 9;12(9):1866. doi: 10.3390/microorganisms12091866.
Networks are widely used to represent relationships between objects, including microorganisms within ecosystems, based on high-throughput sequencing data. However, challenges arise with appropriate statistical algorithms, handling of rare taxa, excess zeros in compositional data, and interpretation. This work introduces a novel Slope-Matrix-Graph (SMG) algorithm to identify microbiome correlations primarily based on slope-based distance calculations. SMG effectively handles any proportion of zeros in compositional data and involves: (1) searching for correlated relationships (e.g., positive and negative directions of changes) based on a "target of interest" within a setting, and (2) quantifying graph changes via slope-based distances between objects. Evaluations on simulated datasets demonstrated SMG's ability to accurately cluster microbes into distinct positive/negative correlation groups, outperforming methods like Bray-Curtis and SparCC in both sensitivity and specificity. Moreover, SMG demonstrated superior accuracy in detecting differential abundance (DA) compared to ZicoSeq and ANCOM-BC2, making it a robust tool for microbiome analysis. A key advantage is SMG's natural capacity to analyze zero-inflated compositional data without transformations. Overall, this simple yet powerful algorithm holds promise for diverse microbiome analysis applications.
基于高通量测序数据，网络被广泛用于表示包括生态系统中的微生物在内的对象之间的关系。然而，在合适的统计算法、稀有分类群的处理、成分数据中的过多零值以及解释方面存在挑战。这项工作引入了一种新颖的斜率矩阵图（SMG）算法，主要基于基于斜率的距离计算来识别微生物组相关性。SMG有效地处理成分数据中任何比例的零值，并且涉及：（1）在一个设定内基于“感兴趣的目标”搜索相关关系（例如，变化的正方向和负方向），以及（2）通过对象之间基于斜率的距离来量化图的变化。对模拟数据集的评估表明，SMG能够将微生物准确地聚类到不同的正/负相关组中，在敏感性和特异性方面均优于Bray-Curtis和SparCC等方法。此外，与ZicoSeq和ANCOM-BC2相比，SMG在检测差异丰度（DA）方面表现出更高的准确性，使其成为微生物组分析的强大工具。一个关键优势是SMG具有在不进行转换的情况下分析零膨胀成分数据的天然能力。总体而言，这种简单而强大的算法在各种微生物组分析应用中具有前景。
