Copy Number Alteration (CNA)

CNAs are inferred from non-uniform sequencing coverage across the genome. In liquid biopsy, the number of aligned reads in a genomic bin correlates with the DNA copy number in the original sample.

• Gains produce elevated read depth relative to a diploid baseline
• Losses manifest as depressed coverage
• Homozygous deletions show near-zero coverage in affected loci
• Detection requires GC bias correction and normalization against a Panel of Normals (PoN) to remove systematic technical artifacts

CNAs are classified by their genomic footprint, which reflects distinct mutational mechanisms and biological consequences.

• Focal amplifications (< 1 Mb) often target specific oncogenes like ERBB2, EGFR, or MYC, driving clonal proliferation
• Arm-level alterations span entire chromosomal arms (e.g., 1q gain, 8p loss) and result from chromosomal instability
• Whole-genome doubling events are identified by broad ploidy shifts across multiple chromosomes
• Focal events are more likely to be actionable therapeutic targets

Circular Binary Segmentation (CBS) is the foundational algorithm for partitioning the genome into regions of constant copy number from noisy read-depth data.

• Recursively tests for change-points using a maximum t-statistic
• Models the data as a series of piecewise constant segments
• Outputs log2 ratio values for each segment relative to a reference ploidy
• Modern alternatives include Hidden Markov Models (HMMs) and Gaussian process regression for improved sensitivity at low tumor fractions

The amplitude of CNA signals in cfDNA is directly proportional to the circulating tumor DNA (ctDNA) fraction. This relationship enables non-invasive tumor burden quantification.

• ichorCNA and similar tools jointly estimate tumor fraction and copy number profiles
• A 10% tumor fraction produces a 0.1-fold deviation from diploid log2 ratios for a heterozygous deletion
• Tumor fraction estimates correlate with RECIST radiographic response and overall survival
• Serial monitoring of CNA amplitude tracks clonal dynamics under therapeutic pressure

Unlike point mutations that require prior knowledge of driver genes, CNAs serve as tissue-agnostic cancer signals detectable across multiple tumor types.

• Over 90% of metastatic cancers harbor at least one somatic CNA
• Copy number instability scores quantify genome-wide CNA burden as a single metric
• Specific CNA patterns (e.g., BRCA1 loss, AR amplification) inform targeted therapy selection
• CNA-based fragmentomics integrates copy number with fragmentation features to enhance detection sensitivity

Copy number alterations frequently co-occur with loss of heterozygosity (LOH), where one parental allele is deleted while the other remains or is amplified.

• Copy-neutral LOH maintains diploid copy number but eliminates allelic diversity
• LOH at tumor suppressor loci (e.g., TP53, PTEN) is a common mechanism of biallelic inactivation
• Allele-specific copy number callers integrate B-allele frequency (BAF) shifts with read-depth signals
• BAF analysis requires heterozygous germline SNP positions as informative markers

The primary computational method for detecting CNAs from sequencing data. Unlike variant calling, which looks for mismatches, read-depth analysis counts the number of aligned sequencing reads in non-overlapping genomic bins. Systematic over- or under-representation of reads relative to a diploid baseline indicates amplifications or deletions, respectively. This method is highly sensitive to GC bias and requires sophisticated normalization.

A critical normalization step for accurate CNA calling. The guanine-cytosine (GC) content of a genomic region non-linearly affects sequencing coverage due to preferential amplification during library preparation. GC bias correction uses locally weighted scatterplot smoothing (LOESS) or similar regression models to model and remove this artifact. Without it, high-GC regions falsely appear amplified and low-GC regions falsely appear deleted.

The study of cfDNA fragmentation patterns to infer tissue-of-origin and epigenetic state. Fragment length analysis reveals that ctDNA is often shorter than non-tumor cfDNA. Nucleosome footprints—regions protected from degradation—map to gene regulatory elements. Integrating fragmentomic features with read-depth signals improves the sensitivity of CNA detection, particularly for low tumor-fraction samples.

The proportion of sequencing reads carrying a specific alteration. In CNA analysis, VAF shifts of heterozygous SNPs within a region of loss-of-heterozygosity (LOH) deviate from the expected 50%. A segment with a VAF approaching 100% suggests a deletion of the alternate allele, while a VAF near 0% indicates deletion of the reference allele. This provides orthogonal evidence for copy number changes.

A reference dataset of sequencing data from healthy individuals used to model technical noise. A Panel of Normals captures recurrent artifacts, polymorphic copy number variants, and systematic mapping biases present in a specific assay. By comparing a tumor sample against this PoN, algorithms can suppress false-positive CNA calls and increase the signal-to-noise ratio for true somatic events.