Computational prediction of replication origins: a comparative review of methods, benchmarks, and trends from heuristics to deep learning

Origins of DNA replication (Ori) are genomic loci where DNA synthesis begins, essential for bacterial viability, eukaryotic development, and genome integrity. While early computational methods relied on compositional skews and motif heuristics, recent years have seen a shift toward artificial intelligence approaches. This review systematically surveys Ori prediction methods published through August 2025, integrating biological background with curated datasets, benchmarks, and a classification of computational strategies from rule-based models to deep learning. We summarize performance across organisms, cell types, and platforms, and highlight key trends, challenges, and opportunities in benchmarking, interpretability, and cross-species generalization-providing both an accessible entry point for practitioners and a roadmap for future methodological development.