Sequence determinant and functional relevance of 8-oxoguanine RNA modification unveiled from foundation-model-based predictor

8-Oxoguanine (o⁸G) is an oxidative RNA modification that plays critical roles in cellular processes including stress responses and RNA integrity regulation. Notably, emerging evidence has implicated o⁸G modifications in cancer progression and development. However, research progress is limited by the lack of efficient high-throughput detection methods and computational tools for analyzing sequence distribution and regulatory mechanisms. Here, we present OBOE (foundation-model-based prediction of 8-oxoguanine sites), the first computational framework for o⁸G site prediction, developed by fine-tuning multiple pre-trained language models (DNABERT, RNABERT, BERT, and BioBERT). Benchmarking experiments demonstrated that OBOE significantly outperforms conventional machine learning methods, validating its superior capability in capturing RNA sequence features and modification patterns. Furthermore, we applied TF-MoDISCo to extract biologically meaningful motifs from model attributions, followed by validation of sequence similarity and enrichment using STREME and TOMTOM analysis. We identified recurrent GC-rich sequence motifs and CTC-like patterns associated with o⁸G modifications, suggesting potential cis -regulatory elements involved in oxidative stress responses. To facilitate model accessibility, we implemented two community resources: (1) a user-friendly web platform for online o⁸G prediction and (2) freely available source code and processed datasets.