Abstract
Acute myeloid leukaemia (AML) is a hematopoietic malignant tumour, whose growth and metastasis have been found to be closely correlated with liquid-liquid phase separation (LLPS), however, the molecular mechanisms and immunological value of LLPS in AML has not been reported. Consequently, this study aims to develop a precise prognostic risk model predicated on LLPS-associated key genes in AML. We analyzed differentially expressed genes (DEGs) from AML vs. control samples (GSE9746), intersected these with LLPS-related genes (LLPSRGs) to identify differentially expressed LLPS-related genes (DE-LLPSRGs) and used univariate Cox regression to find those linked to prognosis. AML subtypes were created through consensus clustering, and DEGs between them were determined. Overlapping prognostic DE-LLPSRGs with inter-subtype DEGs identified candidate genes. We employed Least absolute selection and shrinkage operator (LASSO) to pinpoint three key genes—SLC4A1, SCRN1, and HOPX—for the risk model, which proved effective in assessing AML prognosis. Incorporating risk score, age, and category, a nomogram was developed showing promising potential for clinical utility. Immune analysis revealed variations in certain immune cells across risk groups. Drug sensitivity tests highlighted significant differences in the efficacy of several drugs among these groups. This model integrating the three key geness offers a novel insight into AML prognosis prediction.