Integrated single-cell and bulk transcriptomic profiling reveals coagulation-associated heterogeneity and prognostic SPINT2 in acute myeloid leukemia

Coagulation-associated genes (CARGs) have been implicated in tumor progression and immune regulation, but their cellular heterogeneity and prognostic relevance in acute myeloid leukemia (AML) remain incompletely understood. Here, we sought to characterize the cellular and molecular landscape of CARGs in AML and to identify CARG-associated prognostic determinants with potential functional relevance.

Single-cell RNA sequencing data were first analyzed to characterize CARG activity across distinct cellular populations in AML. Bulk transcriptomic datasets were subsequently integrated to identify CARG-related molecular subtypes and establish a prognostic scoring system using machine-learning approaches. The associations of the CARG prognostic score with biological pathways, immune characteristics, and drug sensitivity were further evaluated. Single-cell expression patterns, cell-cell communication, and computational perturbation analyses were used to investigate key prognostic genes, followed by experimental validation in AML cells.

Single-cell transcriptomic analysis revealed marked heterogeneity of CARG activity across AML cell populations, with particularly high activity observed in erythroid and cycling erythroid populations. Building on these cellular-level observations, bulk transcriptomic analysis identified three distinct CARG-related molecular subtypes with significantly different survival outcomes and biological characteristics. An integrated machine-learning framework established a seven-gene CARG prognostic score that showed prognostic stratification in both the development and independent external validation cohorts. High CARG prognostic scores were associated with distinct metabolic and immune-related transcriptional programs and differential ex vivo responses to several therapeutic agents. Among the model features, SPINT2 showed prominent prognostic importance and displayed cell-type-specific expression at single-cell resolution. Single-cell analysis revealed cell-type-specific SPINT2 expression, while computational perturbation analyses suggested associations with erythroid/heme-related and oxidative metabolic programs. Functional experiments further demonstrated that SPINT2 silencing suppressed AML cell proliferation and promoted apoptosis.

Our study defines a cellular and molecular landscape of CARG heterogeneity in AML and establishes an integrated CARG-based framework for prognostic stratification. The combination of single-cell and bulk transcriptomic analyses, computational perturbation, and functional validation supports SPINT2 as a candidate gene with functional relevance to AML cell growth and survival and provides a basis for further investigation of coagulation-associated biology in AML.