Resolved clinical failures in a high-throughput sequencing workflow.

Identified root cause of clade misclassification and eliminated recurrence in COVID-19 sequencing.

High-throughput sequencing instrument used for clinical viral lineage reporting
Impact
  • Resolved lineage misclassification across a 1,500+ sample sequencing run
  • Identified root cause despite acceptable QC metrics
  • Established new QC thresholds to prevent recurrence
  • Restored confidence in clinical reporting and avoided unnecessary rework

The Problem

A high-throughput SARS-CoV-2 sequencing workflow was producing incorrect lineage assignments in clinical results despite appearing to meet quality control thresholds.

Key Issues:

  • Misclassification observed in both clinical samples and controls
  • Initial concern that cross-sample contamination was driving incorrect results
  • Runs appeared technically acceptable, masking underlying issues
  • Errors skewed toward earlier viral clades (not immediately recognized by the team)

What Was Breaking

The system was not failing in one place. It was failing across the workflow.

  • QC thresholds were met, but did not reflect true signal quality
  • Insufficient targeted-region signal reduced sensitivity
  • Missing lineage-defining mutations biased classification toward earlier clades
  • Upstream input variability propagated into downstream interpretation
  • Contamination was suspected, but not the actual root cause

System-Level Approach

Led end-to-end investigation across run design, RNA input, library prep, sequencing performance, and computational interpretation.

  • Connected upstream biological inputs to downstream lineage assignment
  • Analyzed targeted-region signal relative to classification outcomes
  • Identified bias toward earlier clade assignments driven by insufficient signal
  • Correlated input variability with reduced coverage and mutation sensitivity
  • Ruled out cross-sample contamination as the root cause

Outcomes

  • Identified insufficient targeted-region signal as the root cause
  • Established QC thresholds to prevent lineage miscalls
  • Implemented guardrails for low-signal misinterpretation
  • Updated run acceptance criteria
  • Established a repeatable system-level investigation framework

Why This Worked

The issue was not visible within any single step of the workflow. It emerged from the interaction between biological input, sequencing performance, and computational interpretation.

By analyzing the system end to end, the investigation uncovered a failure mode that standard QC metrics did not detect, allowing the team to correct the issue without unnecessary escalation.

This work clarified assay sensitivity limits, improved robustness, and ensured accurate and reliable clinical interpretation going forward.

Working on something similar?

Start a conversation.