Integrative Transcriptomic Signature of tRNA ‐Derived Fragment Targets and Lactylation‐Related Genes for Prostate Cancer Diagnosis

BACKGROUND: Conventional diagnostic strategies for prostate cancer (PCa) are hindered by suboptimal specificity, frequently resulting in overdiagnosis and unnecessary invasive biopsies. We developed an integrative diagnostic framework based on transcriptomic signatures from tRNA-derived fragments (tRFs), target genes, and lactylation-related genes (LRGs), two emerging regulators linked to PCa progression. METHODS: Bulk RNA-seq data from TCGA-PCa (499 tumors, 52 normals) and five independent GEO cohorts were analyzed. Differentially expressed genes (DEGs) were identified using edgeR and subsequently assessed for functional enrichment. Candidate markers were obtained by intersecting DEGs with predicted tRFs targets and curated LRGs. A least absolute shrinkage and selection operator (LASSO) model was trained and validated across external cohorts. RT-qPCR performed clinical confirmation in 30 prostate tissue samples. RESULTS: Eleven genes overlapped among DEGs, tRF targets, and LRGs, and a seven-gene diagnostic signature (CSRP1, HMGN4, CALM1, NEFL, MSN, MKI67, ARID3A) was established. The model showed robust performance across all validation datasets, with AUC values of 0.728-0.904, exceeding those of prostate-specific antigen (PSA) testing. RT-qPCR supported significant dysregulation of five signature genes in clinical samples, notably decreased NEFL and increased MKI67. Decision curve analysis (DCA) suggested the model could reduce unnecessary biopsies by about 30% at commonly used clinical thresholds. CONCLUSION: This integrative model, incorporating tRFs-associated targets and lactylation-related signatures, demonstrates superior PCa detection accuracy compared to conventional PSA testing. The seven-gene signature offers a biologically informed, clinically applicable tool to improve early diagnosis while potentially reducing unnecessary biopsies.