What question did this study set out to answer?

This research aims to identify prognostic features related to de novo lipogenesis in kidney renal clear cell carcinoma to enhance treatment prediction.

March 28, 2026Open Access

In-silico study of machine learning discovers de novo lipogenesis for predicting prognosis and immunotherapy responses in kidney renal clear cell carcinoma

Key Points

This research aims to identify prognostic features related to de novo lipogenesis in kidney renal clear cell carcinoma to enhance treatment prediction.
Utilized TCGA-KIRC as a training cohort and several datasets for validation.
Constructed prognostic genetic features for de novo lipogenesis using multivariate techniques.
Analyzed correlations between genetic features and tumor characteristics at single-cell and RNA-seq levels.
Validated findings across multiple immunotherapy cohorts.
Conducted functional assays to investigate the role of key gene SLC19A1.
Achieved a composite C-index of 0.796 for prognostic features.
Demonstrated significant association of prognostic features with copy number variation and tumor mutational burden.
Found that high-risk prognostic groups exhibited diminished immunotherapy sensitivity.
Highlighted SLC19A1’s critical role in lipid metabolism and tumor behavior.

Abstract

Currently, there is a lack of precise instruments for predicting the prognosis and treatment efficacy of kidney renal clear cell carcinoma (KIRC), often resulting in delayed diagnosis and suboptimal treatment outcomes. De novo lipogenesis (DNL) has been documented to significantly impact the prognosis and treatment of several cancers and may serve as a biomarker for KIRC treatment and outcomes. We utilized TCGA-KIRC as the training cohort and GSE22541, ematb1980, and emtab3267 as validation cohorts to construct prognostic genetic features (DPGS) for DNL using StepCoxbackward + RSF. We confirmed the independent prognostic value of DPGS through multivariate Cox analysis and integrated DPGS with clinical characteristics to construct comprehensive prognostic nomograms. Furthermore, we analyzed the correlations between DPGS and genomic instability, tumor immune microenvironment (TIME), and immunotherapy sensitivity at both the single-cell and RNA-seq levels. These findings were further validated in the GSE78220, IMvigor210, and Checkmate immunotherapy cohorts. Finally, we used immunohistochemistry (IHC), Western blotting (WB), and cell function assays to elucidate SLC19A1’s critical role in KIRC. The composite C-index for DPGS was 0.796, with hazard ratio (HR) values in multivariate Cox regression analyses of 1.06 and 1.01 in the TCGA-KIRC and EMTAB1980 datasets, respectively. These findings suggest that DPGS can function as an independent prognostic risk factor. Furthermore, the area under the curve (AUC) values for DPGS at 2, 3, and 5 years exceeded 0.67 across four cohorts, indicating robust prognostic predictive performance. In addition, DPGS exhibited a significant association with copy number variation (CNV) (R = 0.33, P < 0.001) and tumor mutational burden (TMB) (R = 0.19, P < 0.001). DPGS was found to accumulate in tumor epithelial cells, cancer-associated fibroblasts (CAFs), and macrophages, and acted on effector CD8 T cells via the HLA-E – KLRC1/NKG2A pathway, thereby diminishing immunotherapy sensitivity in the high-risk DPGS group. Furthermore, knockdown of the key gene SLC19A1 further reduced lipid storage in KIRC and mitigated the malignant phenotype. Our work underscores the critical role of DPGS in modulating TIME, influencing immunotherapy response, and predicting prognosis in KIRC. The integration of DPGS and nomograms based on DNL provides innovated insights into survival prediction and therapy, contributing to the development of DNL-targeted therapeutic strategies.

Mark Helpful

Bookmark

Relay

View Full Paper