Background: Immunogenic cell death (ICD) has emerged as a key factor in cancer prognosis; however, its application in prognostic modeling for acute myeloid leukemia (AML) remains unexplored. Therefore, this study aimed to develop a prognostic model for AML based on ICD-related genes.

Methods: ICD-related genes exhibiting differential expression in AML were identified from The Cancer Genome Atlas (TCGA) database. Initially, candidate genes were filtered using univariate cox proportional hazards model (Cox) and least absolute shrinkage and selection operator (LASSO) Cox regression analyses before being incorporated into a prognostic model. An ICD-related gene signature was then established via the survival package. Furthermore, immune cell infiltration was evaluated using Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data (ESTIMATE), immunedeconv, and ggstatsplot.

Results: Of the total 34 ICD-related genes, 27 demonstrated differential expression, with five genes correlating with AML prognosis. By LASSO regression, we identified four ICD-related genes, including cluster of differentiation 4 (CD4), interleukin-10 (IL10), caspase-1 (CASP1), and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA). Increased risk scores in our model were indicative of a poorer prognosis in AML patients (p < 0.001). Time-dependent Receiver Operating Characteristic (ROC) analysis demonstrated strong predictive capability, with overall survival rates of 0.709, 0.650, and 0.768 at 1, 3, and 5 years, respectively. Nomogram calibration curves confirmed its predictive accuracy. Furthermore, risk scores were substantially correlated with immune cell infiltration (p < 0.05). Moreover, suppression of CASP1 expression significantly diminished AML cell proliferation and increased their sensitivity to chemotherapeutic agents in vitro (p < 0.01).

Conclusion: This study developed a prognostic model for AML using four ICD-related genes (CD4, IL10, CASP1, and PIK3CA). This model demonstrates good predictive accuracy and shows close association with immune infiltration, consistent with functional verification results in vitro, underscoring its potential utility in prognosis assessment and therapeutic decision-making.