Kilianski address the utility of the Oxford Nanopore MinION instrument in species/strain-level identification from three viruses and a bacterial strain through targeted amplicon sequencing. This is a novel and unreported experiment which will be of interest to those wishing to evaluate the MinION for clinical or environmental identification of microbes.

My main issues with this manuscript concern the superficial level of reporting on the MinION sequencing. Given the pace of change in the MinION access programme there have been numerous chemistry changes (R6, R7, R7.3) and library preparation kit changes (SQK-MAP-001 to 004) as well as numerous protocol changes. Judging from the methods description this is likely to be R7 chemistry and SQK-MAP-002 with an additional overnight incubation stage. It is known by the MAP community that the overnight incubation step dramatically reduces yields, which might be one of the reasons for the very low numbers of reads retrieved during this study.

Major Compulsory Revisions

1) Something has gone wrong with these sequencing runs. It may be a fault of the library construction or perhaps supplied reagents. The generation of 296-1335 reads in a 6 hour run is well below expected performance for this device. The authors need to place these results in context with other reports and determine why the performance is so low (bad flowcells? overnight incubation step? other?). Did the authors do a lambda control library to determine whether these results are a result of library preparation? What QC values did these flowcells give? Ideally this would be repeated with latest chemistry (R7.3 and SQK-MAP-004) but if not possible then some additional context is needed here. 2) Improve reporting - what version chemistry, what version library kit, what METRICHOR workflow, what protocol version used? Any deviations from protocols? 3) Are the reads used one-direction (template and/or complement) or 2D? What is the % 2D? 4) Are they categorised as 'normal quality' (# template events > # complement events) or HQ2D (# complement events > #template events) - see Loman 2014 GigaScience for further definition. What % of each? 5) Why did the authors use BLASR for alignment? We found that for R7 normal quality 2D reads that LAST (or BWA MEM -ont2d) gives better results. 6) Was E. coli CS used in library preparation? 7) Where do reads not mapping to target organism come from? 8) "When analyzing the entire genome of E. coli (5-6mb) or poxviruses (150-280kb), the limited data generated after only 6 hours of runtime would make it difficult to identify or characterize organisms because of low coverage depth over a particular stretch of nucleic acids" - we were able to identify E. coli in a matter of minutes using our whole-genome datasets.

Minor essential revisions

9) Hanging sentence "see." in text.

Discretionary revisions

10) Figure 1 heat map is hard to interpret, could the % identities be reported in text?

Level of interest An article of importance in its field Quality of written English Acceptable Statistical review No, the manuscript does not need to be seen by a statistician. Declaration of competing interests I am a participant in the MinION early access programme.