Prediction of Response to Immunotherapy based on scRNA-seq analysis

Immune checkpoint blockade (ICB) therapy has demonstrated remarkable treatment efficacy in a diverse range of cancers. However, it faces the challenge that only a small proportion of patients benefit from it. Using single-cell RNA-sequencing (scRNA-seq) data to predict patients' responses to ICB is a potential strategy for realizing precision medicine in ICB therapy.We set up an accessible pipeline for scRNA-seq data obtaining and analysis in the context of clinical use, and develop a tool named RedeTIL which can analyze single-cell features, including cell abundance and spatial topology of cells, to predict ICB-response. In order to demonstrate the predictive strength of scRNA-seq data, we obtained scRNA-seq data from 11 colorectal cancer (CRC) patients undergoing anti-PD-1 therapy according to our setted pipeline. Besides we collected publicly available scRNA-seq data of 39 breast cancer patients prior to receiving anti-PD-1 therapy, and data derived from 427 patients/samples across 15 tumor types for which objective response rates of anti-PD-1 or anti-PD-L1 were already reported. We used RedeTIL to extract these single-cell-derived features and then correlated them with clinical outcomes.Correlation analysis between single-cell based features and clinical outcomes shows the strong predictive strength of scRNA-seq data. For example, abundance of specific T cell subsets (e,g. CXCL13+ T and PDCD1+ T) has been found to correlate with shrinkage of CRC, and is more enriched in breast cancer responders compared to non-responders. Notably, the infiltration change metrics evaluated by RedeTIL have emerged as the most robust and accurate predictors. The spatial proximity of CD69+ T cells to cancer cells has been shown to correlate with shrinkage of CRC (Pearson R = 0.67, P = 0.02), and is also significantly enriched in Breast cancer responders than non-responders (P = 0.001), moreover, it has exhibited a high correlation with clinical objective response rate to Anti-PD-1/PD-L1 therapy across various cancer types (Pearson R = 0.8, P < 0.001). Additionally, RedeTIL can be utilized to recommend mono- or combined drugs for individual patients.Our results proven that the application of scRNA-seq in predicting the response to immunotherapy in the clinical field is feasible, which can provide abundant and accurate information for treatment decisions, so as to achieve precision therapy.