Development and verification of a deep learning-based m(6)A modification model for clinical prognosis prediction of renal cell carcinoma.

BACKGROUND: The deep learning-based m(6)A modification model for clinical prognosis prediction of patients with renal cell carcinoma (RCC) had not been reported for now. In addition, the important roles of methyltransferase-like 14 (METTL14) in RCC have never been fully explored. METHODS: A high-level neural network based on deep learning algorithm was applied to construct the m(6)A-prognosis model. Western blotting, quantitative real-time PCR, immunohistochemistry and RNA immunoprecipitation were used for biological experimental verifications. RESULTS: The deep learning-based model performs well in predicting the survival status in 5-year follow-up, which also could significantly distinguish the patients with high overall survival risk in two independent patient cohort and a pan-cancer patient cohort. METTL14 deficiency could promote the migration and proliferation of renal cancer cells. In addition, our study also illustrated that METTL14 might participate in the regulation of circRNA in RCC. CONCLUSIONS: In summary, we developed and verified a deep learning-based m(6)A-prognosis model for patients with RCC. We proved that METTL14 deficiency could promote the migration and proliferation of renal cancer cells, which might throw light on the cancer prevention by targeting the METTL14 pathway.