Leveraging diverse liquid-liquid phase separation patterns to predict the prognosis and immunotherapy of pediatric...

Despite improvements in 5-year survival rates for pediatric acute myeloid leukemia (P-AML) over recent decades, the relapse rate remains high. This challenge is compounded by the absence of reliable prognostic biomarkers, which limits the effectiveness of predictive, preventive, and personalized approaches in P-AML management. Emerging evidence has indicated that the aberrant liquid-liquid phase separation (LLPS) can alter the spatiotemporal coordination of biomolecular condensates, thereby contributing to tumorigenesis and progression. However, the role of LLPS in P-AML remains unclear.

A comprehensive multi-omics analysis of LLPS-related genes in P-AML was conducted using both bulk RNA sequencing and single-cell RNA sequencing data. Based on LLPS gene expression profiles, a prognostic risk model was developed through Kaplan-Meier survival analysis, least absolute shrinkage and selection operator (LASSO) regression, stepAIC, and Cox regression analyses. This model aims to predict patient prognosis, immune cell infiltration patterns, responses to immunotherapy, and sensitivity to targeted therapies in P-AML. Additionally, a nomogram integrating the risk model with clinical characteristics was constructed to enhance clinical applicability.

The LLPS-related risk model was developed as an independent prognostic indicator in the TARGET-AML cohort and validated in the external cohorts. Results demonstrated that patients in the low-risk group had significantly better overall survival than those in the high-risk group. In addition, a nomogram was constructed to enhance the clinical applicability of the risk model. Integrated analysis of bulk and single-cell transcriptome data revealed that the LLPS-related signature correlated with cancer hallmarks, immune checkpoint genes, and key components of the tumor microenvironment. Distinct drug sensitivities were observed between risk groups, with P-AML patients exhibiting varied responses to Docetaxel, Paclitaxel, and Sunitinib.

In summary, we developed a robust and effective risk model for predicting prognosis, tumor microenvironment characteristics, and responses to immunotherapy and targeted therapies in P-AML. This model offers valuable insights for advancing personalized and precision medicine approaches in P-AML treatment strategies.