Improvements to the ARTIC multiplex PCR method for SARS-CoV-2 genome sequencing using nanopore

Key results

1. 01

   ShownThe revised library preparations substantially increase the fraction of reads that can be demultiplexed with double-ended barcoding while making barcode misassignment rarer.

   With porechop on libraries of 24 individually barcoded amplicons: demultiplexing rate 42.15% (Baseline), 67.1% (GunIt), 74.5% (LoCost); incorrect assignment 0.070%, 0.010% and 0.002%, a 35x reduction for LoCost over Baseline alongside a 32.4% higher demultiplexing rate. Single-end barcoding stayed worse (0.219% for LoCost). Runs produced about 5M reads in 12 hours. · Protocol improvements; Discussion

2. 02

   ShownSubstituting reagents and shrinking reaction volumes lowers material cost per sample, 96-barcode runs lower it sharply, and a first flowcell wash saves more than any later wash.

   MinION with 24 barcodes: £33.42 (GunIt), £24.91 (LoCost), £16.85 (LoCost with one wash); with 96 barcodes: £18.19, £10.49 and £8.47. PromethION with 96 barcodes, LoCost and two washes gives £10.08. Reagent prices were those listed on Science Warehouse for the University of Birmingham on 1st April 2020. · Run configurations and reagent costs; Discussion; Cost model

3. 03

   ShownOf the primer schemes compared on a cDNA dilution series, V3 gave the most complete coverage over the widest input range and the least uneven coverage, so fewer reads are needed per sample.

   All schemes (V1, V2, V3, Itokawa) reached full coverage at 1e-02 input with 131,072 reads; V3 alone held maximum coverage from 1e-01 to 1e-04 and led at 1e-05, while Itokawa led at 1e-06. V3 had the lowest coefficient of variation from 1e-01 to 1e-05; the discussion puts the V3 requirement at around 100k reads per sample. · Primer schemes; Discussion

4. 04

   ShownWashing and reloading a flowcell with the same barcode set carries over very little of the previous library, making reuse viable for cost saving.

   Three eight-barcode libraries of E. coli amplicons loaded in turn on one FLO-MIN106 flowcell gave 0.01% carry-over after both the first and the second wash. The cost model caps washes at three per flowcell because it ignores pore loss from washing. · Run configurations and reagent costs; Amplicon generation; Library preparation and sequencing

5. 05

   ShownReplacing the post-PCR bead clean-up and the normalisation step with a fixed dilution removes hours of hands-on work without hurting barcode assignment.

   A 1:10 dilution in place of the clean-up left barcode assignment unaffected and saved 1-2 hours; the authors attribute this to PCR product concentration plateauing near 100 ng/ul by cycle 35, which self-normalises input. GunIt held a mean 71.9% demultiplexing rate across 8 clinical runs. · Protocol improvements; Discussion; Library preparation and sequencing