Prognostic marker related to nicotinamide metabolism in ovarian cancer based on multiple machine learning approaches

Nicotinamide metabolism plays a critical role in the formation and progression of ovarian cancer (OC). This study aimed to develop a prognostic marker for OC, focusing specifically on nicotinamide metabolism pathways, using multiple machine learning approaches. Gene expression data from TCGA training cohort (n = 427) and two independent validation cohorts: GSE19829 (n = 28) and ICGC-OV (n = 112), were analyzed. A consensus prognostic model was constructed from a combination of 112 machine learning algorithms. Protein-protein interaction networks, survival analysis, gene set enrichment, mutation profiling, and immune infiltration assessments were performed. Single-cell data from GSE147082 (n = 6)were used for quality control, cell annotation, and developmental trajectory analysis. Our results identify stable prognostic genes associated with OC and developed a risk scoring system that effectively stratifies patients into high- and low-risk groups with significant survival differences (p < 0.05). A final prognostic model base on 23 hub genes demonstrated a robust mean concordance index (C-index) of 0.709. Furthermore, we reveal significant disparities in immune checkpoint markers expression between different risk groups, underscoring potential therapeutic implications. Finally, the model is tested on additional cancer types and clinical utility is assessed using calibration and decision curve analysis. This multiple machine learning tool develops a prognostic marker related to Nicotinamide metabolism for individuals with OC and could potentially be applied to other cancers in the future. Ovarian cancer is a serious disease, and current prediction tools are limited in determining its progressions in different patients. This study examined how a natural substance in the body, called nicotinamide (form of vitamin B3), affects cancer growth. Using publicly available health data and computer-based analysis methods, we develop a tool that identifies 23 genes linked to patient outcomes. This tool reliably separates patients into high-risk and low-risk groups. We also find that immune responses differ between these groups, which may help guide treatment choices. Our findings can help doctors make more personalized treatment decisions and improve survival for ovarian cancer patients. This approach could be applied to other cancers in the future. Yang et al. develop a prognostic marker related to nicotinamide metabolism in ovarian cancer (OC) based on multiple machine learning approaches. Authors identify prognostic genes associated with OC and develop a risk scoring system that effectively stratifies individuals into high- and low-risk groups.