Peer Review #2 of "Rapid multi-locus sequence typing direct from uncorrected long reads using Krocus (v0.1)"

Abstract

Genome sequencing is rapidly being adopted in reference labs and hospitals for bacterial outbreak investigation and diagnostics where time is critical. Seven gene multi-locus sequence typing is a standard tool for broadly classifying samples into sequence types, allowing, in many cases, to rule a sample out of an outbreak, or allowing for general characteristics about a bacterial strain to be inferred. Long read sequencing technologies, such as from Oxford Nanopore, can produce read data within minutes of an experiment starting, unlike short read sequencing technologies which require many hours/days.However, the error rates of raw uncorrected long read data are very high. We present Krocus which can predict a sequence type directly from uncorrected long reads, and which was designed to consume read data as it is produced, providing results in minutes. It is the only tool which can do this from uncorrected long reads. We tested Krocus on over 700 isolates sequenced using long read sequencing technologies from PacBio and Oxford Nanopore. It provides sequence types for isolates on average within 90 seconds, with a sensitivity of 94% and specificity of 97% on real sample data, directly from uncorrected raw sequence reads. The software is written in Python and is available under the open source license GNU GPL version 3. Manuscript to be reviewed Abstract:Genome sequencing is rapidly being adopted in reference labs and hospitals for bacterial outbreak investigation and diagnostics where time is critical. Seven gene multi-locus sequence typing is a standard tool for broadly classifying samples into sequence types, allowing, in many cases, to rule a sample out of an outbreak, or allowing for general characteristics about a bacterial strain to be inferred. Long read sequencing technologies, such as from OPford Nanopore, can produce read data within minutes of an ePperiment starting, unlike short read sequencing technologies which require many hours/days. However, the error rates of raw uncorrected long read data are very high. We present Krocus which can predict a sequence type directly from uncorrected long reads, and which was designed to consume read data as it is produced, providing results in minutes. It is the only tool which can do this from uncorrected long reads. We tested Krocus on over 700 isolates sequenced using long read sequencing technologies from PacBio and OPford Nanopore. It provides sequence types for isolates on average within 90 seconds, with a sensitivity of 94% and specificity of 97% on real sample data, directly from uncorrected raw sequence reads. The software is written in Python and is available under the open source license GNU GPL version 3.