Translating Data Into Clinical Tools: An Integrative Strategy for Precision Biomarker Identification in Soft Tissue Sarcoma Diagnosis and Prognosis.

Abstract

Objectives: Soft tissue sarcomas (STSs) are rare, heterogeneous cancers with over 70 subtypes, often diagnosed late due to diagnostic complexity, leading to poor outcomes. We aimed to identify and validate novel transcriptomic biomarkers for the diagnosis and prognosis of STS using an integrative machine learning and bioinformatics framework applied to publicly available cohorts.

Methods: RNA-seq and clinical data from 261 STS samples were obtained from The Cancer Genome Atlas (TCGA). A multi-step analytical pipeline was implemented, including differential expression analysis, functional enrichment, protein-protein interaction network construction, and clinical correlation assessment. Machine learning algorithms were employed for feature selection and model development. Diagnostic performance was evaluated using receiver operating characteristic curve analysis, and prognostic value was assessed using Kaplan-Meier survival analysis.

Results: We identified a 26-gene prognostic signature significantly associated with overall survival (15 upregulated and 11 downregulated genes). For diagnosis, A1CF alone showed an AUC of 0.70, while the combinations A1CF-ATP6V0D2 and A1CF-LECT2 achieved AUCs of 0.743 and 0.796, respectively. External validation confirmed dysregulation of A1CF in tumor tissues compared with controls.

Conclusions: This study identifies a novel 26-gene prognostic signature and A1CF-based diagnostic panels for STS using computational methods. These biomarkers represent exploratory candidates for future investigation in STS diagnosis and prognosis; however, experimental and prospective clinical validation are required before their potential use in early detection, risk stratification, or personalized management.