Heterogeneity of Acute Myeloid Leukemia patients explored through single-cell and single-sample gene regulatory networks.

Acute myeloid leukaemia (AML) is a genetically heterogeneous disease, driven by diverse mutations and epigenetic alterations that disrupt normal haematopoiesis. Understanding how this heterogeneity manifests at the gene regulatory level can provide insight into disease mechanisms and uncover potential targets for personalized treatment strategies. In this study, single-cell RNA sequencing data from AML patients and healthy controls were used to infer patient gene regulatory networks (GRNs) for progenitor, monocyte, and dendritic cells. Consensus networks were constructed using four inference methods (ARACNE, CLR, MRNET, GENIE3). Additionally, single-sample single-cell networks were constructed using the LIONESS approach for all patients and cell types. Dimensionality reduction applied to both network statistics and adjacency matrices of consensus networks revealed limited clustering, reflecting the biological heterogeneity of AML. Notably, dimensionality reduction and classification models based on single-cell networks achieved distinct patient clustering and perfect discrimination accuracy. This indicates that single-cell GRNs can capture patient-specific signatures of gene regulation. Pathway enrichment analysis of highly connected and predictive genes highlighted distinct regulatory programs across cell types and individuals. These findings demonstrate the potential of GRNs from single-cell data to distinguish between patients and may serve as a foundation for the development of personalized biomarkers and therapeutics.