modFDR: a rigorous method to evaluate the reliability of nanopore sequencing for detecting DNA modifications in real applications

While nanopore sequencing is increasingly used for mapping DNA modifications, it is important to recognize associated false-positive calls, as they can mislead biological interpretations. To assist biologists and methods developers, we describe a framework, modFDR, for rigorous evaluation that emphasizes the use of the false discovery rate with rationally designed negative controls capturing both general background and confounding modifications. Our critical assessment across multiple DNA modifications shows that while nanopore sequencing performs reliably for high-abundance modifications--including 5-methylcytosine (5mC) at CpG sites in mammalian cells and 5-hydroxymethylcytosine (5hmC) in mammalian brain cells--it produces a substantial fraction of false-positive detections for low-abundance modifications, such as 5mC at CpH sites, 5hmC, and N6-methyldeoxyadenine (6mA) in most mammalian cell types. Although newer models improve certain aspects, systematic false positives remain, and we further observe elevated false negatives for 5mCpG when benchmarked against orthogonal enzymatic methods. This study highlights the urgent need to incorporate modFDR into future methods development, evaluation, and biological studies, and advocates prioritizing nanopore sequencing for mapping abundant rather than rare modifications in biomedical applications.