Index Hopping

Index hopping occurs primarily on patterned flow cells using ExAmp chemistry (Illumina). During cluster generation, unincorporated adapter oligos containing free index sequences can prime a neighboring cluster, causing the original sample index to be replaced by a different one. This results in a read that is correctly mapped to the genome but assigned to the wrong sample in the final demultiplexed output.

• Primary cause: Free-floating index primers in the library pool
• Rate: Typically 0.1% to 2% of reads, but can exceed 10% in multiplexed single-cell or low-input libraries
• Exacerbated by: High multiplexing, degraded samples, and over-amplification

Index hopping is a critical confounder in liquid biopsy applications where the biological signal is inherently low. A single hopped read carrying a mutant allele from a high-VAF sample can be misassigned to a negative control or a low-burden sample, generating a false-positive variant call.

• Minimal Residual Disease (MRD): A hopped read at 0.01% VAF can mimic a true relapse signal
• ctDNA monitoring: Contamination distorts longitudinal allele frequency trends
• Single-cell: Hopping between multiplexed cells creates phantom doublets
• Mitigation: Unique dual indices (UDIs) and computational index filtering are mandatory

The primary wet-lab defense against index hopping is the use of Unique Dual Indexing (UDI). Unlike combinatorial dual indexing, where the same i7 and i5 combination may appear in multiple wells, UDI ensures that each sample has a unique, pre-defined pair of i7 and i5 indices.

• Detection: A read with an index pair not in the UDI manifest is flagged as hopped and discarded
• Design: Indices are typically 8 or 10 nucleotides with a minimum edit distance of 3 to prevent misreading
• Limitation: UDIs prevent misassignment but do not recover the lost read; they simply filter the contamination

Bioinformatic methods can rescue data from libraries prepared without UDIs or estimate residual contamination. These tools model the expected cross-contamination rate and probabilistically reassign or down-weight reads.

• DecontX (Celda): Uses a Bayesian hierarchical model to estimate contamination fractions per sample
• Souporcell: Leverages genetic variation to cluster cells and identify cross-sample doublets in single-cell data
• Index-switching filter in Mutect2: Flags read pairs where the molecular barcode suggests a different sample origin
• Post-hoc correction: Subtracts expected hopped allele counts from observed counts using a contamination matrix

Beyond indexing chemistry, careful experimental design minimizes the opportunity for index hopping to confound results. These practices are essential for CLIA-certified liquid biopsy workflows.

• Physical separation: Do not multiplex high-input tumor samples with low-input cfDNA samples on the same flow cell
• Index balancing: Ensure equal representation of all indices to prevent free-adapter excess
• Library quantification: Accurate molarity normalization prevents over-amplification of individual libraries
• Negative controls: Include a no-template control (NTC) in every run to empirically measure the hopping background
• Fresh reagents: Oxidized or aged ExAmp reagents increase the hopping rate

Accurate estimation of the index hopping rate is required for assay validation and regulatory submissions. The rate is calculated by comparing observed cross-contamination between samples with known, orthogonal genotypes.

• PhiX spike-in: Hopping from PhiX into sample libraries is measured by the presence of PhiX reads in sample-demultiplexed FASTQs
• Genotype concordance: Comparing homozygous SNP calls between samples; a heterozygous call in a sample homozygous for the alternate allele indicates hopping
• UMI collision: In UMI-based assays, a UMI family with mixed indices is a direct readout of hopping
• Formula: Hopping Rate = (Misassigned Reads) / (Total Reads Assigned to Recipient Sample)

A random nucleotide barcode ligated to individual DNA molecules before amplification. UMIs enable computational deduplication by tagging each original template molecule, allowing bioinformaticians to distinguish true biological variants from PCR duplicates and index hopping artifacts. In liquid biopsy, UMIs are critical for achieving the molecular consensus required to call rare variants below 0.1% VAF.

The number of unique, non-duplicate DNA molecules in a sequencing library, reflecting the diversity of the original input material. Index hopping artificially inflates apparent complexity by assigning reads from one sample to another's index. True library complexity is measured after computational deduplication using UMIs, and low complexity indicates insufficient input material or over-amplification.

A synthetic nucleotide sequence incorporated into library adapters to uniquely tag individual starting molecules. Unlike sample indexes which identify the sample of origin, molecular barcodes identify each unique DNA fragment. When combined with duplex sequencing, dual molecular barcodes on both strands enable near-perfect error correction, suppressing index hopping cross-talk to below 1 error per 10,000,000 bases.

An error-correction method that independently sequences both strands of a DNA duplex using complementary UMIs. True mutations must appear at the same position on both strands, while PCR errors, oxidative damage, and index hopping artifacts appear on only one strand. This dual-strand consensus approach achieves an error rate of less than 1 in 10^7, making it the gold standard for ctDNA detection.

A bioinformatic strategy leveraging molecular barcodes and redundant sequencing to build consensus sequences. Reads sharing the same UMI are collapsed into a single consensus read, suppressing random polymerase errors and index-switching noise. The method requires a minimum number of reads per UMI family (typically 3-5) to achieve statistical confidence in the corrected base call.

The lowest concentration of analyte reliably distinguishable from background noise. Index hopping directly degrades LoD by introducing false-positive variant reads from high-VAF samples into negative controls. Computational mitigation using UMIs and stringent index quality filtering is essential to maintain an LoD of 0.01% VAF or lower for clinical ctDNA applications.