Наука Просто
OncologyStudy analysis4 min readAugust 17, 2026

cfDNA fragmentation in liquid biopsy creates genomic blind spots

Liquid biopsy does not sample the genome uniformly. Across 317 cfDNA samples, researchers mapped reproducible “penumbra” regions with reduced representation that may weaken signals from clinically relevant cancer loci.

cfDNA fragmentation can create blind spots in liquid biopsy.

Illustration: Nauka Prosto, created with AI assistance.

cfDNA fragmentation in liquid biopsy may hide precisely the genomic regions a test is trying to read when it searches for cancer-associated mutations. In an analysis of 317 samples, the authors found reproducible regions in which circulating DNA was systematically underrepresented — genomic “penumbra” zones where the sequencing signal was consistently weaker.

Cell-free DNA does not circulate as intact chromosomes. It is released from dying cells as short fragments, many of which remain wrapped around nucleosomes, the protein complexes that package DNA. DNA in open chromatin is less protected by nucleosomes and may therefore be more exposed to nucleases, enzymes that cut DNA. That means different parts of the genome may not survive equally well between the cell and the blood sample.

Mapping the genomic penumbra

The researchers developed a computational framework called UPASHAYA and applied it to publicly available low-pass whole-genome cfDNA sequencing data. The analysis included eight cohorts: 54 breast cancer samples, 25 bile duct cancer samples, 28 gastric cancer samples, 78 lung cancer samples, 29 ovarian cancer samples, 35 pancreatic cancer samples, 36 cirrhosis samples and 32 samples from healthy individuals.

They divided the genome into non-overlapping 10-kilobase windows and measured cfDNA representation in each window while accounting for GC content and sequence mappability. Signals were then aggregated within each cohort. Regions with consistently low representation were classified as “penumbra regions.”

About 1.4% of the genome fell within prominent penumbra regions shared across all of the studied cohorts. Another roughly 16% showed more moderate penumbra-like characteristics. The pattern was not evenly distributed across chromosomes: chromosomes 17 and 19, both gene-rich, emerged as major depletion hotspots. The extent of the effect also differed among cohorts. In the lung cancer cohort, for example, about 15% of genomic windows were classified as penumbra regions.

Why some regions lose signal

One of the most informative findings was the relationship with chromatin organization. Across the cancer cohorts, penumbra regions were enriched in open chromatin — genomic regions in which DNA is less tightly protected by nucleosomes and is therefore more accessible.

The fragment characteristics were consistent with increased degradation. Within penumbra regions, short cfDNA fragments under 150 base pairs accounted for about 16–21% of fragments in the cancer cohorts, compared with roughly 10–11% in the healthy and non-cancer cohorts. Fragment ends also showed sequence motifs consistent with preferential nuclease cleavage. These observations do not directly prove that a particular nuclease causes the depletion, but several independent features support a biological fragmentation process rather than a purely technical sequencing artifact.

The effect also did not appear to be explained simply by how much tumor DNA entered the circulation. Similar penumbra enrichment was observed in breast and ovarian cancer cohorts, even though those cancers typically differ substantially in ctDNA shedding. That comparison supports the idea that chromatin architecture itself helps determine which genomic regions are preferentially lost from circulating DNA.

Why this matters for liquid biopsy

A low-coverage region becomes clinically relevant when it contains a mutation a test is intended to detect. The authors found that penumbra regions overlapped known oncogenes and tumor-suppressor genes, as well as exons containing recurrent mutational hotspots. Loci examined in detail included TP53, KRAS and ARID1A.

The study, however, did not measure how many mutations were actually missed by clinical tests, nor did it establish a patient-level false-negative rate. The underlying datasets were generated by low-pass whole-genome sequencing, which limits statistical power at individual hotspot positions. Cohort sizes were uneven, and the same samples were not sequenced at multiple depths to determine directly how increasing coverage changes information loss.

For that reason, the authors position UPASHAYA not as a new diagnostic classifier but as an additional quality-control layer. Mapping regions that are intrinsically vulnerable to cfDNA depletion could help identify loci that deserve more cautious interpretation, deeper sequencing or different panel designs. Clinical implementation would still require independent validation and more standardized pre-analytical conditions.

The central message is that a liquid biopsy does not preserve the genome as a perfectly uniform copy. The biology of cfDNA breakdown can systematically weaken the signal from particular genomic regions, and recognizing those blind spots may improve how cancer-derived DNA is interpreted from blood.