Trained ModiDeC models for site-specific m5C detection by nanopore direct RNA sequencing

This dataset contains trained ModiDeC deep-learning models used in the study"ModiCal: A targeted calibration workflow for site-specific m5C validationby nanopore direct RNA sequencing". Models are provided for two benchmark systems:(i) Saccharomyces cerevisiae 25S rRNA and(ii) Dengue virus genomic RNA (DENV gRNA). For yeast 25S rRNA, two model configurations are included:(a) single-site models trained and calibrated for m5C detection at position C2278, and(b) a dual-site model trained and calibrated for simultaneous, enzyme-dependentdetection of m5C at positions C2278 and C2870. For each configuration, models corresponding to the baseline state, bulkfalse-positive suppression (BulkFP), and final calibrated states are provided.The dual-site calibrated model was used to generate the enzyme-dependent m5Cprofiles shown in Figure 3 of the manuscript. For DENV gRNA, models are provided for the baseline state, bulk false-positivesuppression (BulkFP), and three iterative calibration rounds (cal1–cal3) usedfor site-specific validation at position C1218.