cfDNA Recovery Engineering

ONCOLOGY LIQUID BIOPSY · METHODS RESOURCE

Recover molecules,
not nanograms.

A decision framework for maximizing unique biological cfDNA/ctDNA molecules from plasma—while separating preservation, extraction, library conversion, and sequencing sensitivity.

Model your ceiling Run the screen
01

Protect the numerator. Start with independent endogenous molecules in plasma, then retain them through extraction and library construction. Error correction can reduce false positives; it cannot recreate a molecule that never entered the tube or library.

HARD QUANTITATIVE CEILING

Count what can exist

Mass is a useful process monitor only after contamination is excluded. For sensitivity, locus copies and molecule survival are the physically meaningful currency.

A

Endogenous locus ceiling

Nlocus = Vplasma × clocus

Use a short endogenous ddPCR assay to estimate amplifiable copies per mL. This ceiling is upstream of extraction.

B

Mass proxy

NGE ≈ 303.0 × VmL × Cng/mL

Haploid genome equivalents using 3.3 pg/genome. Invalid as an endogenous cfDNA proxy when leukocyte gDNA contaminates the mass.

C

Expected mutant molecules

λ = Nlocus × VAF × Rext × Rlib

Recovery terms are length- and topology-dependent. Measure them for the fragment class your assay actually detects.

D

Sampling probability

P(X≥k)=1−e−λ Σi=0k−1 λi/i!

The Poisson bound exposes why a nominal LoD is not assured in a specimen containing only a few expected mutant molecules.

E

Distinct-read saturation

E[U] = M(1−e−R/M)

If M unique library molecules receive R read pairs, deeper sequencing asymptotically approaches—not exceeds—M. Spend upstream before buying reads beyond saturation.

INTERACTIVE MODEL

Molecule budget calculator

A planning model, not an LoD validation. Replace defaults with locus-specific ddPCR and measured process recoveries.

Probability of ≥2 mutant molecules—
Locus copies entering extraction—
Expected mutant library molecules—
Mass-derived haploid GE—
Estimated endogenous GE*—
Expected distinct reads—
95% one-hit VAF floor†—

*A sensitivity illustration only; subtraction of an estimated gDNA fraction cannot rescue an invalid mass measurement. †One-hit Poisson floor across the chosen independent loci after extraction and conversion; assay errors and locus dependence are excluded.

MEASUREMENT ARCHITECTURE

Localize loss before optimizing

“Yield” conflates at least five distinct processes. Each needs its own denominator and control.

1

Plasma recovery

Blood volume → cell-free plasma volume. Track tube underfill, additive dilution, hemolysis, buffy-coat carryover, time and spin history.

mL plasma / mL blood
2

Extraction capture

Copies loaded → copies bound. Rebind the discarded supernatant with a fresh matrix under the same condition.

Rbind ≈ 1 − Nflowthrough/Ninput
3

Release

Bound copies → first eluate. Perform a standardized second elution to reveal residual release loss.

Rrelease = (NE1+NE2)/(Ninput−Nflowthrough)
4

Mass yield

Total nucleic-acid signal. Audit HMW gDNA, carrier and assay floor; never treat Qubit alone as the biological endpoint.

ng and fragment profile
5

Library conversion

Extracted copies → unique sequenceable families. Measure by duplex UMI at fixed effective depth.

families / extracted copy
Mandatory recovery audit: spike orthogonal 60–80 bp, 160–180 bp and ~320 bp standards before extraction at non-competitive abundance; quantify input, first eluate, second eluate and second-pass-supernatant eluate. This separates short-fragment capture, release and size bias rather than hiding them in one percent-recovery number.

TUBE CHEMISTRY

Disclosed ≠ inferred ≠ proprietary

Commercial names do not disclose a formulation. Current IFUs/SDS define what is known; patents define research space, not the contents of a purchased tube.

Tube / armWhat current documents disclosePatent or independent evidenceEngineering implication
Plain EDTABD K2EDTA tube: 18 mg per 10 mL draw (1.8 mg/mL intended blood).Processing delay can raise cfDNA through leukocyte contamination; same-day rapid separation remains the clean baseline.Best chemically transparent control. Optimize elapsed time, fill, first spin, buffy-coat clearance and plasma recovery before extraction chemistry.
Streck Cell-Free DNA BCTCurrent IFU: K3EDTA plus liquid cell preservative; current SDS: <0.22 mL, pH 5.6, composition withheld as trade secret. Do not add or dilute components.A Streck-assigned patent teaches imidazolidinyl or diazolidinyl urea + EDTA + glycine quencher; its stock-solution claim space includes 20–80% urea derivative, 1–10% quencher and 1–20% EDTA, with ~50–400 µL per tube. Röth et al. detected formaldehyde crosslinks in PBMCs from tested Streck tubes.Do not label the current commercial formula “formaldehyde” from a patent or biomarker study. Follow the RUO digestion requirement: proteinase K ≥30 mAU/mL with chaotrope, 60°C for 1 h.
PAXgene Blood ccfDNARUO IFU: 10 mL blood + 1.5 mL additive; non-crosslinking, without formaldehyde/formaldehyde releasers. SDS leaves composition proprietary.Patent exemplars combine K2EDTA, Q-VD-OPh caspase inhibitor, PEG300, higher-MW PEG and DMPA/butanamide. One exemplar reports 5 µM Q-VD-OPh, 0.5% HMW PEG and 1% DMPA. Its narrative states 5.5% v/v PEG300, but 402.5 µL in 1.5 mL additive + 10 mL blood calculates to ≈3.5% v/v: an internal patent-text inconsistency.Use the exact mixture fractions: additive is 1.5/11.5 = 13.0% and blood is diluted to 10/11.5 = 0.870×. Treat patent concentrations only as mechanistic DOE arms—not a reverse-engineered product recipe.
PEG–salt preservative
PATENT RUO ARM		US20230365961A1 “Preservative A”: 33% w/w PEG (PEG8000 preferred), 5% NaCl, 2% EDTA, 0.023% sodium azide; ~1.5 mL into ~10 mL total gives ≈4.95%, 0.75%, 0.30%, 0.00345%, respectively.Patent examples report 28–30 day stabilization; no independent validation is assumed here.A distinct non-aldehyde research arm. Compare with/without azide; sodium azide is hazardous and may constrain enzymes, disposal and downstream multiomics. Not a clinical or commercial formulation.

Tube studies are lot-, workflow-, temperature- and assay-specific. A stable total-cfDNA value can still mask changes in size distribution, damaged templates, or target-specific amplifiability.

EXTRACTION PHYSICAL CHEMISTRY

Engineer each unit operation

A “magnetic” method is a handling format, not a binding chemistry. Surface, crowding, salt, pH, alcohol, kinetics and elution jointly determine the captured fragment population.

01

Proteolyze & release

Proteinase K and detergent disrupt protein complexes; chaotrope denatures proteins and suppresses nucleases. Fixed QIAamp reference: for 4 mL plasma, 400 µL PK + 3.2 mL ACL containing 1.0 µg carrier RNA, 60°C for 30 min.

Preservative-bound material may require longer digestion: Streck RUO specifies ≥30 mAU/mL PK at 60°C for 1 h with chaotrope.

02

Adsorb selectively

High ionic strength, reduced water activity and charge screening promote DNA adsorption to silica; alcohol strengthens dehydration. Adsorption is multiphasic and context-dependent—not one universal “salt bridge.”

Historical Boom L6: 5.25 M GuSCN, 50 mM Tris pH 6.4, 20 mM EDTA, 1.3% Triton X-100. It is an open reference, not QIAamp’s formula.

03

Wash without losing

Remove proteins, inhibitors, chaotrope and alcohol while DNA remains bound. Over-drying can impair release; inadequate drying transfers ethanol and depresses ligation or PCR.

Measure inhibition by spike-after-extraction ΔCq and record dry time, not merely wash count.

04

Elute efficiently

Low-ionic-strength buffer near pH 8 rehydrates the surface and DNA. Smaller elution raises concentration but can lower total recovery; prewarming, contact time and a second elution reveal release-limited conditions.

Report total copies recovered and eluate concentration separately.

Commercially disclosed concentration windows

QIAamp DSP ACL / ACB30–<50% w/w GuSCN
QIAamp DSP ACW130–<60% w/w guanidine HCl
QIAamp DSP proteinase K1–<3% w/w

SDS concentration bands are hazard disclosures, not complete recipes and not necessarily final reaction concentrations.

SURFACE & FRAGMENT BIAS

Choose for the molecule class

The same plasma can yield different quantities, size distributions and mutant-molecule counts across workflows. Validate recovery by length and topology.

SiO₂

Silica membrane

Mature large-volume workflows; high surface area and vacuum/centrifugal handling. In a 14-lab round robin, bead workflows averaged 29% fewer total cfDNA and 41% fewer mutant molecules than silica—but this is a workflow observation, not a law of bead surfaces.

  • Watch column capacity and viscosity
  • Test ultrashort capture explicitly
  • Second-pass supernatant localizes non-binding
Si

Silica-coated magnetic

Silane surfaces pair silica-like binding with automation. Dynabeads MyOne Silane are 1 µm and 40 mg/mL; the vendor gDNA protocol uses 20 mg/mL PK plus guanidine and isopropanol.

  • Starting reference, not a cfDNA protocol
  • Titrate surface loading and mixing
  • Avoid declaring equivalence to membranes
COO−

Carboxylated PEG–salt

SPRI uses carboxylated particles with crowding agent and salt. BOMB reference buffer: 2.5 M NaCl, 20% PEG8000, 10 mM Tris pH 8, 1 mM EDTA, 0.05% Tween-20, plus 2% v/v bead stock.

  • Excellent cleanup/size-selection platform
  • Not automatically optimal for crude plasma
  • PEG/salt ratio creates size bias
PAA

Other charged surfaces

Poly(acrylic acid)-coated magnetic particles demonstrate that surface charge, polymer density and salt response can be engineered independently of magnetic handling.

  • Demand plasma-matrix validation
  • Measure inhibition and carryover
  • Benchmark against a silica control
±

Lysine-modified zwitterionic

He et al.’s MSP-ZEWB arm uses ampholytic ion exchange with low nonspecific protein adsorption. Reported condition: 300 µL plasma, 2.0 mg beads, glycine-HCl adsorption buffer with final 0.8% PEG8000, 25°C for 10 min; two 60% ethanol washes; 40 µL 25 mM Tris-HCl pH 8.9 elution.

  • High-upside ultrashort research arm
  • Only 24 colon-cancer comparisons
  • Patented/proprietary bead preparation; control kit was not run at its instructed volume
Ø

Extraction-free / direct-to-library

WO2020106893A1 bypasses purification and reports greater recovery of unique and sub-100 bp cell-free fragments by generating libraries directly from plasma.

  • Eliminates binding and elution losses
  • Transfers EDTA, protein, lipid and hemolysis inhibitors into library chemistry
  • Evidence is patent examples centered on microbial cfDNA—not independently validated oncology ctDNA performance

The ultrashort blind spot

Plasma contains ~50 bp ultrashort cfDNA and ultrashort ssDNA detectable with high-affinity extraction and single-stranded library preparation. A high recovery for a 170–180 bp spike does not prove recovery of 50–80 bp molecules.

Design rule: include a 60–80 bp endogenous assay plus multi-size spike-ins. Do not size-select away the analyte before deciding whether it is informative for the intended assay.

Carrier and blocking chemistry

Carrier RNA can improve low-input capture by saturating nonspecific loss sites; QIAamp’s reference uses 1.0 µg per 4 mL reaction. It also inflates nucleic-acid readouts and conflicts with cfRNA/multiomic endpoints.

Design rule: use low-bind plastics as a fixed condition. If testing carrier, quantify by locus-specific assays; avoid carrier DNA in untargeted sequencing, where it consumes reads and can contaminate inference.

SCREENING BLUEPRINT

A fractional-factorial recovery screen

Use a resolution IV 27−3 design: 16 factorial runs with generators E=ABC, F=BCD, G=ACD. Main effects are clear of two-factor interactions; two-factor interactions are aliased and require confirmation.

16

factorial runs

+4

center points

+4

current-SOP controls

=24

runs per surface / pool

AGuSCN3.5 ↔ 5.25 M3.5 M proposed; 5.25 M Boom reference
BFinal alcohol30 ↔ 50% v/vproposed DOE start points
CPK digestion30 ↔ 60 min, 60°CQIAamp / Streck documented anchors
DCarrier RNA0.5 ↔ 1.5 µgproposed around 1.0 µg official baseline
EBinding pH5.5 ↔ 7.0proposed DOE start points
FSurface loading0.5× ↔ 1.0× capacityproposed, vendor-rated basis
GElution volume20 ↔ 60 µLinside QIAamp’s 20–150 µL range

Block silica membrane and silica-coated bead separately. Do not pool their effects as a single “silica” response.

Stage 1

Controlled discovery

Pooled plasma; randomized runs; ≥3 independent extraction days as blocks. Every run includes blank, spike-only control, first eluate, standardized second elution and second-pass binding of the discarded supernatant.

Stage 2

Biological verification

Carry the two best robust conditions forward in paired testing across 8 donors × 3 technical replicates. Model donor as a random effect; preserve disease-relevant plasma if available.

Selection

Optimize the lower bound

Primary: duplex-UMI consensus families per input mL at fixed effective depth. Co-primary process measure: endogenous 60–80 bp ddPCR copies/mL. Select on confidence-bound performance, not maximum mean Qubit.

Secondary endpoints160/320 bp spike recoveryHMW gDNA ratioΔCq inhibitionfragment profileresidual chaotrope / ethanolhands-on time

FIT FOR PURPOSE

“Best recovery” depends on the assay

Lock the downstream measurement before optimizing extraction; otherwise the winning condition may enrich molecules the assay cannot use.

Targeted SNV / indel

UMIs, short inserts and locus copies matter. Ultrashort molecules may not span both primers or capture probes. Control HMW gDNA and validate template amplifiability.

Tumor-informed MRD

Volume × loci × unique conversion drives sampling. Duplex error suppression improves specificity, while multiple independent loci raise the chance of sampling a true mutant molecule.

Fragmentomics / WGS

Native ends and length distribution are signal. Avoid hard size selection, carrier DNA and re-fragmentation; evaluate single-stranded library preparation for ultrashort fractions.

Methylation

Bisulfite or enzymatic conversion adds loss and sequence bias. Include a post-conversion control and evaluate carrier compatibility with the conversion and quantification method.

cfRNA / multiomics

Carrier RNA invalidates RNA mass and may consume library capacity. Preservative, protease and nuclease choices must be jointly qualified across analytes.

CNV / aneuploidy

Leukocyte gDNA changes copy-number baseline and fragment composition. Control plasma separation and HMW fraction before interpreting greater DNA mass as better input.

AUDITABLE EVIDENCE LEDGER

Discovery → primary record

Scholarly records were discovered and fetched through the official Consensus connector, then linked to DOI records. Formulation facts come from current IFUs/SDS, patents or official protocols and are labeled by evidence class.

Consensus-fetched paperOfficial protocol / IFU / SDSPatent exemplar—not product formula
  1. Deveson et al., 2021. Analytical validity of ctDNA sequencing assays. DOI · Consensus
  2. Risberg et al., 2018. Collection and processing effects. DOI · Consensus
  3. Medina Diaz et al., 2023. Streck BCT preanalytics. DOI · Consensus
  4. van der Leest et al., 2022. Dutch national round robin. DOI · Consensus
  5. Polatoglou et al., 2022. Extraction method comparison. DOI · Consensus
  6. Sorber et al., 2017. cfDNA isolation kit comparison. DOI · Consensus
  7. Lehle et al., 2023. Automated extraction techniques. DOI · Consensus
  8. Sandberg et al., 2025. Variability and spike-in recovery. DOI · Consensus
  9. Devonshire et al., 2014. Standardisation of cfDNA measurement. DOI · Consensus
  10. Hudecova et al., 2022. Ultrashort plasma cfDNA. DOI · Consensus
  11. Newman et al., 2016. Integrated digital error suppression. DOI · Consensus
  12. Katevatis et al., 2017. Low-concentration DNA recovery on silica. DOI · Consensus
  13. Boom et al., 1990. Guanidinium thiocyanate–silica purification. DOI · Consensus
  14. Vandeventer et al., 2012. Multiphasic DNA adsorption to silica. DOI · Consensus
  15. Shaw et al., 2009. Carrier RNA and silica extraction. DOI · Consensus
  16. Bali et al., 2023. Poly(acrylic acid)-coated magnetic particles. DOI · Consensus
  17. Oberacker et al., 2019. BOMB open magnetic-bead platform. DOI · Consensus
  18. Warton et al., 2017. Streck and PAXgene stabilized blood. DOI · Consensus
  19. Cheng et al., 2022. Ultrashort single-stranded plasma DNA. DOI · Consensus
  20. Röth et al., 2023. Formaldehyde crosslinks in PBMCs from Streck tubes. DOI · Consensus
  21. He et al., 2024. Ampholytic ion-exchange magnetic beads for short cfDNA. DOI · Consensus
  22. NCI. Biospecimen Evidence-Based Practices: cfDNA.
  23. BD. K2EDTA 10 mL tube product record.
  24. Streck. Cell-Free DNA BCT IVD IFU · RUO IFU · SDS.
  25. Streck-assigned patent. WO2013123030A2. Candidate chemistry; not evidence of the current product formula.
  26. PreAnalytiX. PAXgene Blood ccfDNA RUO IFU · SDS.
  27. PAXgene-family patents. WO2015140218A1 · WO2017085321A1. Patent exemplars; not marketed formula.
  28. QIAGEN. QIAamp DSP Circulating NA handbook · QIAamp Circulating NA handbook.
  29. Beckman Coulter. AMPure XP protocol · SPRI technology.
  30. Thermo Fisher Scientific. Dynabeads MyOne Silane genomic DNA manual.
  31. Promega. Maxwell RSC Rapid ccfDNA manual.
  32. US20230365961A1. PEG–salt preservative patent exemplar. Research arm only; not a commercial formula or independent validation.
  33. WO2020106893A1. Direct-to-library methods patent. Patent proof of concept, largely microbial cfDNA; not independent oncology validation.
Evidence method. Consensus was used as the scholarly discovery and record-fetch layer. Claims about reagent identity or concentration were then constrained to primary articles, patent text, current manufacturer instructions, safety data sheets and official protocols. Patents demonstrate disclosed embodiments, not commercial composition, clinical validity or freedom to operate.

SCOPE & SAFETY

Research design resource—not clinical advice

This material supports research-method development. It is not a clinical protocol, diagnostic claim, validated medical recommendation, or instruction to alter regulated workflows. Reagents including guanidinium salts, detergents, proteinase K and sodium azide require institutional risk assessment, compatible waste handling and qualified personnel. Confirm all current manufacturer instructions before use.