Oral Presentation GENEMAPPERS 2024

The long and the short of establishing a clinical nanopore sequencing facility for acute care in New Zealand (#46)

Denis M Nyaga 1 , Peter Tsai 1 , Sophie Farrow 1 , Hui Hui Phua 1 , Jing Rong 1 , Clare Gebbie 1 , Polona Le Quesne Stabej 1 , Eric Thorstensen 1 , Gergely Toldi 2 , Kim Gamet 3 , Patrick Yap 3 , Mark Greenslade 4 , Jodi Van Dyk 1 , Justin M O'Sullivan 1
  1. The University of Auckland, Grafton, AUCKLAND, New Zealand
  2. Starship Child Health, Auckland City Hospital Te Whatu Ora Te Toka Tumai, Auckland, New Zealand
  3. Genetic Health Service New Zealand-Northern Hub, Te Toka Tumai, Auckland, New Zealand
  4. LabPLUS, Auckland District Health Board , Auckland, New Zealand

There are ~200 children in high dependency neonatal/paediatric acute care in New Zealand at any one time, requiring a scalable distributed solution for acute care genomics. We have established an expandable acute care clinical pipeline based around the PromethION 2 solo system with connection to a Bayesian AI-based clinical decision support tool Fabric GEM™. In the establishment phase, we have performed benchmarking using Global Alliance for Genomics and Health (GA4GH) benchmarking tools and Genome in a Bottle HG002 - HG007. Evaluating calls for ~3.3x106 truth single nucleotide variants (SNVs) and ~500x103 small insertions-deletion (indels) at read depths of between 24-42X coverage identified SNV recalls = 0.992 ± 0.001, precision = 0.997 ± 0.0006, and F1 = 0.995 ± 0.0008 over a minimum of two sequencing runs completed by different technicians and analysts. Small indel identification approached recalls = 0.838 ± 0.043, precision = 0.922 ± 0.019, and F1 = 0.874 ± 0.032 over the same runs.  Subsequent analyses indicated that the observed variation in recall, precision and F1 was largely limited to correct copies of falsely duplicated regions and areas of collapsed errors with clusters of CHM13 heterozygous (hets) in GRCh38. Rarefaction analyses up to 80X coverage identified that SNV identification plateaus at ~20X coverage, while indel identification plateaus at ~40X coverage.  Analyses of the Coriell Copy Number Variation Reference Panel 1 (CNVPANEL01) demonstrated that large scale genomic variations are reliably detected after ~2M reads, equivalent to ~2hr sequencing time. Application of the pipeline in acute care genomic diagnosis is ongoing. We present the preliminary results from the pipeline validation phase, performed in parallel with established international accredited facilities available to New Zealand’s clinicians.