BACKGROUND

The growth and metastasis of pancreatic cancer (PC) has been found to be closely associated with liquid-liquid phase separation (LLPS). This study sought to identify LLPS-related biomarkers in PC to construct a robust prognostic model.

METHODS

Transcriptomic data and clinical information related to PC were retrieved from publicly accessible databases. The PC-related data set was subjected to differential expression, Mendelian randomization (MR), univariate Cox, and least absolute selection and shrinkage operator analyses to identify biomarkers. Using the biomarkers, we subsequently constructed a risk model, identified the independent prognostic factors of PC, established a nomogram, and conducted an immune analysis.

RESULTS

The study identified four genes linked with an increased risk of PC; that is, , and . Conversely, , and were found to provide protection against PC. These findings contributed significantly to the development of a highly precise risk model in which risk, age, and pathology N stage were categorized as independent factors in predicting the prognosis of PC patients. Using these factors, a nomogram was established to predict survival outcomes accurately. An immune analysis revealed varying levels of eosinophils, gamma delta T cells, and other immune cells between the distinct risk groups. The high-risk patients exhibited increased potential for immune escape, while the low-risk patients showed a higher response to immunotherapy.

CONCLUSIONS

Six genes were identified as having potential causal relationships with PC. These genes were integrated into a prognostic risk model, thereby serving as unique prognostic signatures. Our findings provide novel insights into predicting the prognosis of PC patients.