Freshwater monitoring by nanopore sequencing

Lara Urban; Andre Holzer; J Jotautas Baronas; Michael Hall; Philipp Braeuninger-Weimer; Michael J Scherm; Daniel J Kunz; Surangi N Perera; Daniel E Martin-Herranz; Edward T Tipper; Susannah J Salter; Maximilian R Stammnitz

Minimap2 performed second best at classifying the mock community (lowest RMSE), while also delivering freshwater bacterial profiles in line with the majority vote of other classification tools (Extended Data Figure 1d e), in addition to providing rapid speed (data not shown). Yet, the application of this software to our entire dataset caused insufficient memory errors (at ~150 Gb RAM with kmer length 12), likely due to major sequence redundancies within the SILVA v.132 reference fasta file. Therefore, to run each of our full samples within a reasonable memory limit of 50 Gb, it was necessary to reduce the number of threads to 1, raise the kmer size ('k') to 15 and set the minibatch size ('-K') to 25M (i.e., the number of query bases that are processed at any time), prolonging the runtime of several samples to ~three days.