Path Signatures Enable Model-Free Mapping of RNA Modifications

The introduction of a model-free computational method for RNA modification detection marks a significant advancement in the field of epitranscriptomics. Traditional sequencing methods often introduce biases and fragment RNA, complicating the mapping of modifications. In contrast, nanopore direct RNA sequencing preserves the native structure of RNA, allowing for single-molecule resolution. The new method reframes modification detection as an anomaly detection problem, relying solely on canonical RNA reads. This innovative approach has been validated on densely modified E. coli rRNA, demonstrating its effectiveness in identifying diverse modification types without prior training. By extracting robust features from the raw ionic current signal and computing anomaly scores, this method offers a promising alternative for researchers aiming to map RNA modifications accurately, thereby enhancing our understanding of gene regulation and cellular functions.