ERBB2 as a Prognostic Biomarker in Prostate Cancer: Integration of Single-Cell Transcriptomics, Deep Learning, and...

Biochemical recurrence (BCR) is a critical factor affecting the prognosis of prostate cancer (PCa) patients, while T cell exhaustion and metastatic prostate cancer (mPC)-related genes play significant roles in tumor progression. This study aims to identify key genes associated with BCR by integrating single-cell transcriptomics and deep learning techniques, and to validate their clinical significance. First, we identified highly expressed genes in CD8⁺ T cell exhaustion clusters from single-cell RNA sequencing data (scRNA-seq) of PCa and intersected them with mPC-related genes. Based on these genes, significant prognostic factors were screened using Cox regression analysis, and a deep learning neural network model was constructed to predict the risk of biochemical recurrence in prostate cancer patients. The tumor-infiltrating lymphocyte (TILs) infiltration was predicted by stratifying patients into high- and low-risk groups. ERBB2 was identified as the most predictive gene through model analysis. Subsequently, ERBB2 expression was validated in an independent PCa cohort using immunohistochemistry (IHC), and its association with biochemical recurrence and clinicopathological features was evaluated through survival analysis and statistical methods. The deep learning model demonstrated excellent performance in predicting BCR, with ERBB2 identified as the most important predictive factor. IHC results revealed that patients with high ERBB2 expression had significantly shorter biochemical recurrence-free survival (bRFS) (P < 0.05). Moreover, high ERBB2 expression was significantly associated with higher prostate-specific antigen (PSA) levels, Node-Metastasis (NM) stage, and International Society of Urological Pathology (ISUP) grade (P < 0.05). This study, for the first time, integrates single-cell transcriptomics, deep learning, and IHC to reveal the critical role of ERBB2 in biochemical recurrence of PCa. High ERBB2 expression is not only a potential biomarker for poor prognosis in PCa patients but may also provide a novel target for personalized therapy.