m6A-MAC: evaluation results and figure source data

This dataset provides selected aggregate evaluation results and numerical figure source data for the study "m6A-MAC: target-centered multiscale modeling for RNA m6A site prediction." The archive includes model- and run-level performance metrics, paired model comparisons, retrospective candidate-recovery results, methylation-level ranking analyses, and structural-context and sequence-sensitivity summaries. The files are organized into results and figure-source directories, accompanied by documentation, provenance information, SHA-256 checksums, and a data-partitioning utility. Candidate-level sequences, methylation labels, per-site prediction scores, raw experimental measurements, pretrained encoder weights, and trained model checkpoints are not included. Original datasets, including m6AConquer and GEO resources, should be obtained from their respective providers under applicable terms. The internal development-evaluation set informed model development and is not an independent confirmatory test set. The associated project repository is available at https://github.com/Juseong03/m6A-MAC.