A multi-omics features-based approach integrating immunogenicity and inflammation enhances immunotherapy benefit in clear cell renal cell carcinoma

Yanfeng Xue¹Feng Han¹Shuqing Wei¹Yiqun Zhang²Nan Wang³Ling Liu⁴Zhen Chen⁵Zhihua Pei^6*Hailong Hao^1*

• ¹Cancer Hospital Affiliated to Shanxi Medical University/Shanxi Province Cancer Hospital/ Shanxi Hospital Affiliated to Cancer Hospital, Chinese Academy of Medical Sciences, Taiyuan, China
• ²Frontier Science Center for Stem Cell Research, School of Life Sciences and Technology, Tongji University, Shanghai, China
• ³School of Medicine, Xiamen University, Xiamen, Fujian, China
• ⁴School of Bioinformatics Science and Technology, Harbin Medical University, Harbin, Jilin Province, China
• ⁵The Second Affiliated Hospital of Fujian Medical University, Quanzhou, Fujian Province, China
• ⁶Hubei Key Laboratory of Agricultural Bioinformatics, College of Informatics, Huazhong Agricultural University, Huazhong Agricultural University, Wuhan, China

Abstract Background: Programmed cell death 1 (PD-1) or PD-ligand 1 (PD-L1) blocker-based strategies have improved the survival outcomes of clear cell renal cell carcinomas (ccRCCs) in recent years, but only a small number of patients have benefited from them. Methods: In this study, we developed a multi-omics machine learning (ML) model based on inflammatory and immune signatures (TIs) to predict the response and survival of ccRCC patients to immune checkpoint blockade (ICB) therapy. The research collected RNA-seq and single-cell RNA-seq (scRNA-seq) data from more than 1,900 patients with autoimmune nephropathy and analyzed the genomic and transcriptome profiles of ccRCC patients. The predictive power of the method was validated in more than 1,000 ccRCC patients treated with ICB, and compared to single biomarkers (e.g., PD-L1 expression, TMB). Results: Inflammatory signaling was found to be strongly associated with ICB outcome, and 716 inflammation-related genes were identified that are enriched in the "lymphocyte activation regulation" pathway. The findings suggested that ccRCC patients can be categorized into two subtypes with different treatment responses and prognosis by these features. In addition, the TIs-ML model exhibited superior predictive capabilities compared to an individual biomarker (AUC >0.997) across multiple independent datasets. It demonstrated the capacity to accurately differentiate between responders and non-responders. Furthermore, the model performed more effectively than existing genetic models and functional scores in predicting survival. Conclusion: We propose a TIs-ML prediction model based on multi-omics features that can effectively predict ICB treatment response in ccRCC patients. The model integrates inflammatory and immune features, and its high generalization ability was validated in multiple cohorts. Overall, the TIs-ML approach provides a novel method for guiding precise immunotherapy in ccRCC.