Pan-Cancer transcriptomics reveals diverse R-loop events based on multiple machine learning algorithms

Abstract

Abstract Background: R-loops are three-stranded RNA-DNA hybrids which play an important role in various cellular and chromosomal function including transcriptional regulation and genome instability. Due to the absence of approach characterizing R-Loops events at a larger scale, we developed a computational metrics as R-Loops score to decipher the landscape of R-Loops events at a pan-cancer level and within LUAD patients. Methods: We developed a computational metric, R-Loops Score (RS), to provide a quantified approach of evaluating the R-Loop events. Then, Unsupervised clustering of diverse R-Loops pattern with LUAD patients revealed intratumoral heterogeneity, on the basis of which did we construct a prognostic model and corresponding R-Loops Related Score (RRS) via the multi-machine learning framework for the prediction of clinical outcome. Results: Comparing with patients with low RS, the high RS group had significantly lower survival, higher detectable chromosomal instability (CIN), alongside various oncogenic pathway activities. LUAD patients with lower RS exhibited distinct immune infiltration pattern, better clinical outcomes, and different mutation landscapes. Conclusion: The RS could function as a quantified method to evaluate R-Loops events across individual cancer types. The RRS provided the LUAD patients with a R-Loops-based prognostic model indicating how CIN involves cancerous peculiarities and immune patterns.