Question 7 of 9 Intermediate Troubleshooting
What evidence would lead you to suspect index hopping rather than a genuine rare variant?
Short answer
Suspect index hopping when rare reads carry an otherwise implausible sample assignment pattern linked to a high-abundance library on the same run. Read count alone cannot prove the mechanism.
Why
Multiplexed reads can be misassigned through index switching, sample cross-contamination, demultiplexing mistakes or barcode errors. Unique dual indexes can reveal unexpected i5/i7 combinations and help reduce certain misassignment modes, while molecular evidence is needed to distinguish a genuine low-frequency allele.
How to work it through
- Inspect dual-index combinations, unexpected index-pair counts, shared variants across co-pooled samples, abundance correlation, negative controls and lane or run history.
- Check allele-supporting read positions, mapping quality, molecular families, strand agreement and whether the candidate appears in an independently prepared library.
- Audit sample sheet versions and demultiplexing rules.
An example
A rare allele appears as a handful of reads in a negative control and several low-input samples only on runs containing one heavily positive sample. Invalid index pairs and abundance scaling suggest cross-assignment rather than biological recurrence.
Where the answer stops holding
Unique dual indexes mitigate detectable hopped combinations but do not eliminate carryover before indexing. Tumor-in-normal contamination and clonal hematopoiesis may create genuine low-level signal with different implications.
A strong answer covers
- Suspect index hopping when rare reads carry an otherwise implausible sample assignment pattern linked to a high-abundance library on the same run
- Read count alone cannot prove the mechanism
- Inspect dual-index combinations, unexpected index-pair counts, shared variants across co-pooled samples, abundance correlation, negative controls and lane or run history
Common mistakes
- Calling every low-VAF read index hopping
- Treating zero in a negative control as proof of zero contamination
- Relying on VAF without unique-molecule counts
The interviewer might ask next
How would unique molecular identifiers change your validation?