Nanopore sequencing technologies are used widely in genomics research and their adoption continues to accelerate. ‘Basecalling’ is an essential step in the nanopore sequencing workflow, during which raw electrical signals are translated into nucleotide sequences. The current state-of-the-art basecaller, Oxford Nanopore Technologies (ONT) software ‘Dorado,’ relies on proprietary, platform-specific NVIDIA GPU optimisations bundled in the closed-source ‘Koi’ library. As a result, practical, high-speed basecalling is effectively restricted to a narrow class of supported hardware, limiting accessibility, portability, and innovation. We present (1) ‘Openfish,’ an open-source GPU-accelerated nanopore basecaller decoding library that provides a competitive alternative to ONT’s proprietary Koi library; and (2) Slorado, a fully open-source basecalling framework that supports both DNA and RNA with equivalent accuracy to Dorado. Together, Openfish and Slorado remove the hardware lock-in that currently limits high-performance nanopore basecalling. Our framework scales efficiently across heterogeneous computing environments, from low-power embedded devices to GPU-equipped datacenters, without sacrificing speed or accuracy. Openfish and Slorado are available as free open-source packages for basecalling research, optimisation and deployment beyond the constraints of proprietary software and hardware ecosystems: Openfish: https://github.com/warp9seq/openfish, Slorado: https://github.com/BonsonW/slorado.
Depression subtype classification from social media posts: few-shot prompting vs. fine-tuning of large language models
BackgroundSocial media provides timely proxy signals of mental health, but reliable tweet-level classification of depression subtypes remains challenging due to short, noisy text, overlapping symptomatology,