Common Issues — Failed Mira QC

Low-coverage / Incomplete segment coverage

  • Not enough reads in your library → Back to the lab
    • Ct ≤ 28?
    • Gel image resolves segment bands cleanly?
    • Proper sample number on your flow cell?
  • MIRA ≤ v2.0.0?
    • Reads are being subsampled

Tip

A low Ct usually means more template — start your troubleshooting there before re-running the pipeline.

Common Issues — Failed Mira QC

Minor variant count > 10

  • In standard clinical samples, we have never observed > 6 minor alleles (≥ 5% frequency); the majority have 1–3 per segment
  • Cell cultures regularly have high genetic variability which can result in high counts
  • Could be a co-infection — an unlucky person got both H1N1 and H3N2 at the same time
  • Most likely contamination!

Warning

A pattern of high minor-variant counts across all segments is a strong red flag for contamination.

Example: Per-segment QC results

sample_id total_reads reads_mapped reference % ref. cov. median cov. minor SNVs ≥ 5% pass / fail reason
95f48e8a 95,828 10,388 A_HA_H1 99.82 735 2 Pass
95f48e8a 95,828 5,818 A_MP 100 686 148 minor variants > 10
95f48e8a 95,828 9,528 A_NA_N1 99.79 814 3 Pass
95f48e8a 95,828 9,146 A_NP 100 722 274 minor variants > 10
95f48e8a 95,828 5,704 A_NS 97.1 777 147 minor variants > 10
95f48e8a 95,828 11,790 A_PA 100 607 361 minor variants > 10
95f48e8a 95,828 13,230 A_PB1 100 687 246 minor variants > 10
95f48e8a 95,828 12,175 A_PB2 100 590 391 minor variants > 10
c282a097 16,170 1,992 A_HA_H3 99.82 158 35 minor variants > 10
c282a097 16,170 1,382 A_MP 100 187 12 minor variants > 10
c282a097 16,170 1,852 A_NA_N2 100 178 25 minor variants > 10
c282a097 16,170 1,720 A_NP 100 157 18 minor variants > 10
c282a097 16,170 1,336 A_NS 97.1 206 10 Pass
c282a097 16,170 2,516 A_PA 100 148 22 minor variants > 10
c282a097 16,170 2,712 A_PB1 100 161 23 minor variants > 10
c282a097 16,170 2,660 A_PB2 100 160 28 minor variants > 10

Two samples (95f48e8a, c282a097) — note how nearly every segment fails with Count of minor variants at or over 5% > 10. pass_qc column (= total_reads) omitted for clarity.

Two populations are present

  • One matching the reference (gray)
  • One sharing all the colored SNVs

2+ mutations on one molecule are “phased” or “linked”

A clean bimodal signal across many reads strongly suggests mixed populations / contamination, not random sequencing error.

IGV view of A_NP showing two read populations: a gray reference-matching set and a set sharing colored SNVs

Common Issues — Failed Mira QC

Premature stop-codon?

  • ONT homopolymer issue — may require manual correction
  • DI particle–induced alignment can create the same artifact

Note

Always inspect the alignment around the premature stop before discarding the segment — many premature stops are fixable artifacts, not biology.

DI Particles

  • DI Particles can interfere with NGS
  • Common in polymerase segments
  • Coverage shape can spike at the end — “bat ears”
  • Can create erroneous indel mutations at coverage drop-off points

DI RNA schematic (full-length vRNA vs DI RNA)

Coverage depth plot showing bat-ear spikes near segment ends

Defective interfering particles — Biotechnology Journal, 2014

DI Particles — Alignment View

IGV alignment of A_PB2 showing read pile-up and coverage dropoff with arrowed indel artifact caused by DI particles

Coverage drop-offs (red triangle) align with apparent indels — often a DI-particle artifact rather than a true mutation.

Frameshifts

  • Prevalent in homopolymer regions of Oxford Nanopore sequencing
  • DAIS-Ribosome is frameshift-tolerant
    • Shows as a ~ mutation
  • Convert back to nucleotide space and add or remove a base to fix

Jalview screenshot showing manual edit of homopolymer region in an HA alignment to correct a frameshift

Mira Demands High-Quality Results

  • Together we can stop the “Garbage In” side of the data-analysis mantra: “Garbage In, Garbage Out”
  • Built-in thresholds are for standardizing QC
  • Amended consensus gives us extra information from a single sequence
    • A mixed site may be under active or balancing selection in the host
  • Consider each sample as a whole — if HA and NA pass but all other segments fail, consider why
  • High-priority samples can be useful even when QC thresholds are not met

Quality data in → quality decisions out. Every NIC matters.

Thank You

Questions & Discussion

Ben Rambo-Martin, PhD
Lead Bioinformatics Scientist
US CDC — Influenza Division