Targeted Error Correction

The final analytical gate applies a statistical cutoff to distinguish signal from residual noise. The limit of detection (LoD) is determined by:

• Library complexity: The number of unique input molecules, which sets the theoretical sensitivity floor
• Sequencing depth: The total coverage at the target locus after deduplication
• Local background error rate: The empirically observed noise level at the specific genomic position A variant must exceed a minimum number of supporting molecular families (not just reads) and achieve a VAF significantly above the local background (typically via a Poisson or beta-binomial model). For duplex-consensus workflows, thresholds as low as 0.01% VAF are achievable, enabling detection of a single mutant molecule among 10,000 wild-type copies.

A random nucleotide barcode ligated to individual DNA molecules before amplification. UMIs tag each original template molecule, enabling computational deduplication after PCR. By grouping reads sharing the same UMI, a consensus sequence is built that cancels out random polymerase errors introduced during library preparation. This converts a noisy analog signal into a digital counting problem, enabling absolute quantification of rare variants.

An error-correction method that independently sequences both strands of a DNA duplex using complementary UMIs. True mutations are expected to appear as complementary base pairs on both strands, while single-strand errors from oxidative damage or polymerase mistakes are confined to one strand. By requiring bidirectional concordance, duplex sequencing achieves error rates as low as 10⁻⁷, making it the gold standard for ultra-rare variant detection.

The percentage of sequencing reads at a specific genomic locus that contain a variant allele. In liquid biopsy, tumor-derived variants often present at VAFs below 0.1%, well within the error rate of standard sequencing. Targeted error correction suppresses background noise to lower the limit of detection, enabling reliable variant calling at these ultra-low frequencies where the true biological signal would otherwise be indistinguishable from sequencer noise.

A synthetic nucleotide sequence incorporated into library adapters to uniquely tag individual starting molecules. Unlike sample indices that identify which patient a read belongs to, molecular barcodes identify which original DNA molecule a read came from. This enables:

• Deduplication: Collapsing PCR duplicates into a single consensus read
• Error suppression: Distinguishing true variants from amplification artifacts
• Absolute quantification: Counting input molecules rather than amplified copies

A specialized algorithm designed to distinguish low-frequency true somatic mutations from germline variants, sequencing errors, and mapping artifacts. In targeted error correction workflows, the variant caller operates on consensus reads rather than raw reads, dramatically reducing the false positive rate. Modern callers integrate base quality recalibration, strand bias filters, and position-specific error models to achieve clinical-grade sensitivity for liquid biopsy applications.

The lowest concentration of analyte that can be reliably distinguished from background noise. For liquid biopsy assays, LoD is typically expressed as the minimum VAF detectable with 95% confidence. Targeted error correction directly improves LoD by suppressing the technical noise floor created by polymerase errors (typically ~10⁻³) and sequencer errors (~10⁻²), pushing detection limits down to 0.01% VAF or lower when combined with sufficient input DNA and deep sequencing coverage.