Mike Vella, Senior Director of Machine Learning for Oxford Nanopore Technologies, provided an update on Dorado at the Nanopore Community Meeting in Boston. Vella explained that Dorado “powers Nanopore base-calling” by converting the raw signals into sequences. Stand-alone Dorado is a command-line tool. Dorado is also integrated into MinKNOW. Vella announced that among the updates, there are more basecalling models and improved accuracy. Barcode demultiplexing has also been improved. Dorado has poly(A) support for tail length estimation, plasmids, and interrupted tails. SUP basecalling speed has improved. Vella did mention that most people should use HAC for their applications. However, Vella acknowledged that people will want to use the improved SUP accuracy despite its slower speed. Read correction was improved by adding Herro. Read correction is recommended for assembly but has not been tested for other applications. Improvements in modified base calling are expected. Vela described their vision for designing end-to-end tools for users. I’m glad they continue updating basecalling and how it is integrated into MinKNOW and other tools.
