Deletion‑aware RNA nanopore basecalling with student model training
The project tackles high deletion errors in direct RNA nanopore sequencing by providing a student model trained with QW‑CTC and multi‑scale temporal adapters. It ships training, inference, and statistical analysis scripts that operate on external signal data and pseudo‑labels. Targeted at bioinformatics researchers and developers needing a customizable basecalling pipeline, it adds teacher‑relative evaluation and deletion‑aware training beyond generic basecallers. The implementation is a concrete, reproducible tool rather than just a research idea.
View on GitHub →meizisong888/basecall-rna-deletion