Tailoring iterative decoding schemes to exploit soft information from raw nanopore reads

Develop iterative decoding schemes from coding theory that directly exploit soft information derived from raw nanopore sequencing signals—such as the per-sample state probability matrices produced by CTC- or EDHMM-based basecallers—within integrated basecaller–decoder pipelines like SynDe, and determine their performance and computational complexity for DNA data storage readout.

Background

The paper introduces SynDe, a syndrome-trellis–based, beam-search decoder that integrates error-correcting code constraints directly into nanopore basecalling, using soft information from neural network outputs over raw signals. This approach shows competitive-to-superior performance with substantially lower complexity than prior basecaller-integrated decoders.

While SynDe focuses on convolutional and marker codes, the authors note that broader families of iterative decoding methods from coding theory have not yet been adapted to leverage the rich soft information available in raw nanopore reads. Designing such decoders and integrating them with basecalling could extend error-correction capability and efficiency beyond the specific schemes evaluated here.