The Problem
A high-throughput SARS-CoV-2 sequencing workflow was producing recurring incorrect lineage assignments in a subset of each batch. The initial concern was cross-sample contamination, since misclassification appeared in both clinical samples and controls.
The deeper risk was more concerning: the workflow could appear technically acceptable while producing unreliable downstream interpretation. Standard QC metrics passed, but they did not reflect whether there was enough targeted-region signal to support accurate lineage calls. The failure mode was not visible within any single workflow step. It emerged across the interaction between upstream input variability, assay sensitivity, sequencing signal quality, QC thresholds, and computational interpretation.
- Existing QC thresholds passed despite incorrect viral lineage assignments
- Run-level quality metrics created false confidence in result quality
- Contamination was suspected but was not the true failure mode
Integration & Risk Assessment Approach
Led an end-to-end investigation across run design, RNA input, library prep, sequencing performance, QC thresholds, and computational interpretation. The work focused on finding the failure mode hidden between system layers, not overcorrecting the wrong component.
"The work focused on finding the failure mode hidden between system layers, not overcorrecting the wrong component."
- Evaluated the workflow across biological input, assay performance, sequencing signal, and lineage interpretation
- Connected upstream input variability to downstream classification behavior
- Identified insufficient targeted-region signal as the mechanism biasing calls toward earlier clades
- Ruled out cross-sample contamination as the root cause
- Assessed existing QC thresholds against true interpretive reliability requirements
- Established stronger run-acceptance criteria and interpretation guardrails
Outcomes
- Identified the mechanism driving lineage misclassification and ruled out contamination
- Established QC thresholds calibrated to targeted-region signal quality
- Updated run acceptance criteria for high-throughput clinical reporting
- Restored confidence in clinical reporting without unnecessary workflow redesign
- Created a repeatable framework for investigating cross-system workflow failures
Why This Worked
The issue was not contained within one component. It emerged from the interaction between biological input, assay sensitivity, sequencing signal, QC logic, and computational interpretation. Standard QC metrics could not detect it because the failure existed between layers, not inside one.
An end-to-end assessment identified the hidden failure mode and strengthened the right control points instead of overcorrecting the wrong part of the system. This avoided tens of thousands of dollars in unnecessary customer site visits, troubleshooting runs, and automated workflow changes that would not have addressed the actual root cause.
Beyond the immediate fix, the investigation produced a repeatable framework for diagnosing cross-system workflow failures. The team could apply the same structured approach to future investigations independently, without external support.