@francoisfleuret @bozavlado I suspect it has to with the very local and time-axis dependence of basecalling. That is, a base 10 positions upstream from the read base does not influence the read base’s identity. Found this interesting benchmarking paper tha
2/6 The data is a problem generally discussed in a recent Nature editorial: https://t.co/FUi39Rdo3w, and more specifically in the case of #nanopore sequencing here: https://t.co/owzkEsvtCY.
RT @BioDecoded: Comprehensive benchmark and architectural analysis of deep learning models for nanopore sequencing basecalling | Genome Bio…
RT @BioDecoded: Comprehensive benchmark and architectural analysis of deep learning models for nanopore sequencing basecalling | Genome Bio…