ctDNA Library Prep and Enrichment: What Limits Sensitivity
Team Dynamic Matrices | 5 October 2026
Circulating tumour DNA (ctDNA) is a small fraction of cell-free DNA, often well under one percent. A ctDNA assay is therefore limited by three things: how many mutant molecules you start with, how many of them you convert into sequenceable library, and how well you can tell real variants from errors. This guide covers what happens after extraction. For blood collection, plasma preparation and extraction, see our cfDNA and ctDNA sample prep guide.
We make LASSO, a bead-free capture chemistry that can be used in enrichment steps. We describe where it fits at the end and keep the rest vendor-neutral.
Start with the sampling limit
The first limit is physical, not chemical. A haploid human genome weighs about 3.3 pg, so 10 ng of cfDNA holds roughly 3,000 haploid genome equivalents. At a variant allele fraction of 0.1 percent, that is about three mutant molecules in the whole sample. Even with perfect library conversion and no errors, Poisson sampling means that around one sample in twenty would contain none. Plasma cfDNA yield in healthy donors is typically 1 to 10 ng per millilitre, so input volume, not kit chemistry, sets the floor for the lowest allele fraction you can claim to detect. For broader context on ctDNA analysis and its limits, see Wan et al. (2017) and Heitzer et al. (2019).
Library prep for short, scarce fragments
- Use chemistry built for low input and short fragments. Kits adapted from sheared genomic DNA protocols lose the short end of the cfDNA distribution, which peaks at about 166 bp (Snyder et al., 2016).
- Conversion efficiency matters more than amplification. A molecule lost during end repair or adapter ligation cannot be recovered by PCR. Every percentage point of conversion is a percentage point of sensitivity at low input.
- Consider single-stranded library prep for the shortest fragments. Burnham et al. (2016) found that single-stranded library prep recovered a greater share of sub-100 bp nuclear cfDNA, and more mitochondrial and microbial cfDNA, than double-stranded prep.
- Keep PCR cycles low. Over-amplification increases duplicates and the chance that a polymerase error is called as a variant.
- Add a recovery control at extraction so you can tell loss during sample prep from a true absence of signal.
Unique molecular identifiers and duplex consensus
Raw sequencing and PCR error rates are high relative to a 0.1 percent allele fraction, so low-fraction variants cannot be called from raw reads alone. Two approaches suppress error:
- Unique molecular identifiers (UMIs). Each original molecule is tagged with a short random barcode before amplification. Reads that share a barcode are grouped into a family and collapsed to a consensus, so an error present in only some reads of the family is discarded (Kinde et al., PNAS, 2011).
- Duplex consensus. Both strands of the original duplex are tagged, and a variant is called only if it is seen on both strands. Schmitt et al. (PNAS, 2012) reported a theoretical background error rate below one artefactual mutation per billion nucleotides sequenced.
The trade-off is depth and input. Consensus calling needs several reads per molecule, and duplex calling needs both strands of the same molecule to be recovered and sequenced, so efficiency per input molecule is lower. UMIs are standard for ctDNA at low allele fractions; they matter less for high-burden samples at several percent.
Hybrid capture or amplicon enrichment
| Hybrid capture | Amplicon (multiplex PCR) | |
|---|---|---|
| How it works | Biotinylated probes pull target regions out of the library | Multiplex PCR amplifies the target regions directly |
| Panel size | Suits larger panels and broad coverage | Suits small, focused panels |
| Short fragments | Works with short fragments | Each amplicon must fit inside a fragment, and cfDNA peaks near 166 bp, so amplicons must be short |
| Workflow | Longer, with hybridisation and wash steps | Shorter and simpler |
| Main risks | Off-target co-capture and losses at nanogram input | Primer bias and PCR errors early in the workflow |
| Example | CAPP-Seq, a hybrid-capture design for lung cancer that detected ctDNA in all stage II to IV patients and about half of stage I patients (Newman et al., 2014) | Focused hotspot panels |
Where capture loses signal
In hybrid capture, a sub-one-percent signal is most easily lost at three points: off-target co-capture of genomic fragments that share partial homology with the probes; harsh elution from streptavidin beads, which short fragments tolerate poorly; and surface losses, because at nanogram input adsorption to beads and tubes takes a disproportionate share of a scarce sample. Our cfDNA guide covers these in more detail.
Where sequence-selective capture without beads fits
LASSO capture was designed to run under mild, native conditions with no solid surface: catcher strands on a soluble polymer hybridise to target fragments, the network condenses into a pellet, and release is by strand displacement in the same buffer. In our hands this keeps recovery high and off-target low at the input levels cfDNA workflows operate at. We are running long-fragment genomic DNA capture feasibility work with a clinical lab partner using the same chemistry. If you are designing a custom enrichment, LASSOflex lets you design your own catcher strands, and our multiplexed capture panels post covers panel design.
QC gates for ctDNA sequencing
- Before library prep: fluorometric yield, fragment size distribution and the genomic DNA contamination ratio, as pass/fail gates on every sample.
- After library prep: library complexity (unique molecules versus reads) and duplication rate. If complexity is low, you are input-limited rather than depth-limited.
- After capture: on-target rate and coverage uniformity across the panel.
- Controls: reference material with known allele fractions to measure your real limit of detection, and healthy-donor plasma to measure background error.
References
- Wan JCM et al. Liquid biopsies come of age: towards implementation of circulating tumour DNA. Nat Rev Cancer 17, 223–238 (2017).
- Heitzer E, Haque IS, Roberts CES, Speicher MR. Current and future perspectives of liquid biopsies in genomics-driven oncology. Nat Rev Genet 20, 71–88 (2019).
- Newman AM et al. An ultrasensitive method for quantitating circulating tumor DNA with broad patient coverage. Nat Med 20, 548–554 (2014).
- Kinde I, Wu J, Papadopoulos N, Kinzler KW, Vogelstein B. Detection and quantification of rare mutations with massively parallel sequencing. PNAS 108, 9530–9535 (2011).
- Schmitt MW, Kennedy SR, Salk JJ, Fox EJ, Hiatt JB, Loeb LA. Detection of ultra-rare mutations by next-generation sequencing. PNAS 109, 14508–14513 (2012).
- Burnham P et al. Single-stranded DNA library preparation uncovers the origin and diversity of ultrashort cell-free DNA in plasma. Sci Rep 6, 27859 (2016).
- Snyder MW, Kircher M, Hill AJ, Daza RM, Shendure J. Cell-free DNA comprises an in vivo nucleosome footprint that informs its tissues-of-origin. Cell 164, 57–68 (2016).
- Dynamic Matrices. LASSO: sequence-selective biomolecule isolation by programmable polymer phase separation. Angew Chem Int Ed (2025).
FAQs
It depends on the allele fraction you need to detect. 10 ng is roughly 3,000 haploid genome equivalents, so a 0.1 percent variant is about three molecules, and lower fractions need more input or you accept lower sensitivity. Many library prep kits work down to a few nanograms, but input sets the limit of detection, not the kit.
For variants below about one percent allele fraction, yes: UMIs let you collapse reads from the same original molecule and discard polymerase and sequencing errors. For high-burden samples at several percent, they matter less.
Amplicon enrichment is simpler and faster for small panels, but amplicons must be shorter than the roughly 166 bp cfDNA fragments. Hybrid capture suits larger panels and short fragments, with a longer workflow and more loss at nanogram input.
Cell-free DNA (cfDNA) is all the DNA circulating in plasma, released mainly by apoptosis and largely from blood cells. Circulating tumour DNA (ctDNA) is the tumour-derived fraction of it, often well under one percent of the total.