Machine learning-derived prognostic signature integrating programmed cell death and mitochondrial function in renal clear cell carcinoma: identification of PIF1 as a novel target

• Published in: Cancer Immunology, Immunotherapy : CII Vol. 74; no. 4; pp. 113 - 16
• Main Authors: Cheng, Guangyang, Zhou, Zhaokai, Li, Shiqi, Peng, Fu, Yang, Shuai, Ren, Chuanchuan
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 25.02.2025; Springer Nature B.V; Springer
• Subjects: Apoptosis; Biomarkers, Tumor - genetics; Biomarkers, Tumor - metabolism; Cancer; Cancer Research; Carcinoma, Renal Cell - genetics; Carcinoma, Renal Cell - metabolism; Carcinoma, Renal Cell - mortality; Carcinoma, Renal Cell - pathology; Cell death; Cell Line, Tumor; Correlation analysis; Female; Gene Expression Regulation, Neoplastic; Humans; Immune checkpoint; Immunology; Kidney cancer; Kidney clear cell carcinoma; Kidney Neoplasms - genetics; Kidney Neoplasms - metabolism; Kidney Neoplasms - mortality; Kidney Neoplasms - pathology; Learning algorithms; Machine Learning; Male; Medical prognosis; Medicine; Medicine & Public Health; Mitochondria; Mitochondria - genetics; Mitochondria - metabolism; Mitochondria function; Oncology; Patients; PIF1; Prognosis; Prognostic signature; Programmed cell death; Renal cell carcinoma; Renal function; Survival analysis; Tumor Microenvironment; Prognostic signature; PIF1; Programmed cell death; Kidney clear cell carcinoma; Mitochondria function; Machine learning
• ISSN: 1432-0851; 0340-7004; 1432-0851
• DOI: 10.1007/s00262-025-03967-8

Background The pathogenesis and progression of renal cell carcinoma (RCC) involve complex programmed cell death (PCD) processes. As the powerhouse of the cell, mitochondria can influence cell death mechanisms. However, the prognostic significance of the interplay between mitochondrial function (MF) and PCD remains unclear. Methods We collected sets of genes related to PCD and MF. Using a powerful machine learning algorithm framework, we investigated the relationship between MF and PCD in different cohorts of patients and developed a machine learning-derived prognostic signature (mpMLDPS) related to MF and PCD. Finally, the most appropriate prognostic markers for RCC were screened by survival analysis and clinical correlation analysis, and the effects on renal cancer cells were analysed in vitro. Results mpMLDPS was significantly correlated with the prognosis of RCC patients, and the prognosis was worse in the high mpMLDPS group, and this result was also validated in external independent cohorts. There were associations between mpMLDPS and immune checkpoints, tumour microenvironment, somatic mutations, and drug sensitivity. Finally, a novel RCC prognostic marker PIF1 was identified in model genes. The knockdown of PIF1 in vitro inhibited the progression of renal carcinoma cells. Conclusion mpMLDPS has great potential to serve as a reliable clinical signature to improve the accuracy and reliability of prognostic assessment in RCC patients, thereby choosing the appropriate therapeutic regimen in clinical practice. PIF1 is also expected to be a novel target for the clinical treatment of RCC.