Genome-wide association studies in combination with single-cell genomic atlases can provide insights into the mechanisms of disease-causal genetic variation. However, identification of disease-relevant or trait-relevant cell types, states and trajectories is often hampered by sparsity and noise, particularly in the analysis of single-cell epigenomic data. To overcome these challenges, we present SCAVENGE, a computational algorithm that uses network propagation to map causal variants to their relevant cellular context at single-cell resolution. We demonstrate how SCAVENGE can help identify key biological mechanisms underlying human genetic variation, applying the method to blood traits at distinct stages of human hematopoiesis, to monocyte subsets that increase the risk for severe Coronavirus Disease 2019 (COVID-19) and to intermediate lymphocyte developmental states that predispose to acute leukemia. Our approach not only provides a framework for enabling variant-to-function insights at single-cell resolution but also suggests a more general strategy for maximizing the inferences that can be made using single-cell genomic data.

全基因组关联研究与单细胞基因组图谱相结合，可以深入了解疾病致病遗传变异的机制。然而，疾病相关或性状相关细胞类型、状态和轨迹的识别通常受到稀疏性和噪声的阻碍，特别是在单细胞表观基因组数据的分析中。为了克服这些挑战，我们提出了 SCAVENGE，这是一种计算算法，它使用网络传播以单细胞分辨率将因果变异映射到它们的相关细胞环境。我们展示了 SCAVENGE 如何帮助识别人类遗传变异背后的关键生物学机制，将该方法应用于人类造血不同阶段的血液特征，增加 2019 年严重冠状病毒病 (COVID-19) 风险的单核细胞亚群和易患急性白血病的中间淋巴细胞发育状态。我们的方法不仅提供了一个框架，可以在单细胞分辨率下实现从变异到功能的洞察力，而且还提出了一种更通用的策略，可以最大限度地利用单细胞基因组数据进行推断。