PROFESSOR'S EXPERIMENTAL ATLAS · ONCOLOGY PLASMA

The molecule is
the unit of truth.

A bench-ready program for recovering more unique endogenous cfDNA—from blood tube to duplex-UMI family—without mistaking leukocyte DNA, concentration, or PCR duplicates for gain.

Open the DOE Compare chemistries

Asta MCP status

The configured official endpoint was attempted on 25 Aug 2026 and returned HTTP 403 because no x-api-key is configured.

Discovery fallback

Official Ai2 Semantic Scholar corpus/API surfaces were used as the Asta-family discovery layer, followed by direct primary-source verification. [A][S2]

Verification rule

Formulations below come from papers, patents or current IFU/SDS documents; proposed DOE ranges are visibly labeled.

I · DEFINE THE MOLECULE BUDGET

Every loss must have a denominator.

Carry all outcomes back to original plasma volume. A yield claim without a fragment length, locus, and process boundary is not interpretable.

GENOME INPUT

N_GE ≈ 303 · V_mL · C_ng/mL

Planning estimate using 3.3 pg per haploid genome equivalent. It does not correct for damaged, single-stranded or non-amplifiable DNA.

MUTANT TEMPLATES

λ_mut = N_locus · R_ext · R_lib · VAF

Use endogenous locus copies—not bulk nanograms—when estimating how many independent mutant molecules can exist.

POISSON FLOOR

P(X ≥ k) = 1 − e^−λ Σ_i=0…k−1 λⁱ/i!

No sequencer can detect a molecule that was absent from the library aliquot. Report the probability for the assay's required confirmations.

LIBRARY SATURATION

E[U] = M(1 − e^−R/M)

With M unique templates and R informative read pairs, extra reads asymptotically stop discovering new molecules.

PrimaryEndogenous 60–80 bp ddPCR copies / mL

Assay two independent single-copy loci; add a 160–180 bp assay and a ≥300 bp assay to expose fragment transfer and leukocyte gDNA.

Process accountingFirst eluate + second eluate + rebound supernatant

Measure where molecules went. A second-pass capture of the first supernatant separates nonbinding from irreversible bead loss.

NGS truthDuplex-UMI families / mL at fixed depth

Also report singleton UMIs, family-size distribution, on-target unique depth, insert size, duplication and molecular saturation. [17]

II · CONTROL THE SPECIMEN CHEMISTRY

Commercial label, patent embodiment, and inference are three different things.

Run commercial tubes exactly as their current IFU requires. Patent examples are useful mechanistic hypotheses—not permission to alter a collection device.

Tube	Commercially disclosed	Patent-family evidence	Likely experimental question	Do not infer
K₂EDTAOPEN CHEMISTRY	Standard comparator: about 1.8 mg K₂EDTA per mL whole blood in the cited patent control. EDTA chelates divalent cations but does not stabilize leukocyte membranes.	No proprietary stabilizer is needed to interpret the control.	Immediate processing versus 2 h and 4 h; paired temperature block. Measure short copies/mL and long/short ratio.	More total DNA after delay is not better recovery. NCI targets ≤2 h and treats ≤4 h at room temperature as an alternative. [1]
Streck Cell-Free DNA BCTPROPRIETARY	Current IFU discloses K₃EDTA plus a proprietary liquid preservative and tube-specific handling limits. [2]	Streck-assigned WO2013123030 describes diazolidinyl or imidazolidinyl urea, EDTA and glycine to quench free formaldehyde. [3]	Does preservation alter proteinase demand, apparent short-copy recovery, or library end accessibility versus time-matched EDTA?	A patent claim is not the current batch formula. Do not back-calculate the marketed recipe or add reagents to the tube.
PAXgene Blood ccfDNAPROPRIETARY	10 mL draw into 1.5 mL proprietary non-crosslinking additive; eight inversions. The additive is 1.5/11.5 = 13.0% of the filled-tube volume and the blood dilution factor is 10/11.5 = 0.870. The SDS does not enumerate a complete formula. [4][5]	PAXgene-family patents test aldehyde-free mixtures containing Q-VD-OPh, K₂EDTA, PEG300, high-MW PEG and DMPA. Example 16 lists, per 1.5 mL additive, 57.5 mg PEG10k, 402.5 µL PEG300, 115 µL DMPA, 152 mg K₂EDTA and 11.5 µL Q-VD-OPh stock, with water to volume. [6][6b]	Does the non-crosslinking design improve enzyme accessibility or ultrashort transfer after storage?	Patent inconsistency: 402.5 µL PEG300 divided by the 11.5 mL filled-tube volume gives 3.50% v/v, matching the example's component label, while a later patent narrative states 5.5%. Both appear in the same example. Neither figure is a qualified commercial-formulation claim.

III · ENGINEER THE CAPTURE STEP

Two bead surfaces. Two thermodynamic regimes.

Keep lysis/protein liberation separate from solid-phase capture in the design matrix. Otherwise a “bead effect” can actually be a digestion effect.

ROUTE A

Direct-plasma silica magnetic beads

Chaotrope/alcohol-mediated adsorption to silica-like surfaces. Best developed open anchor: UltraPrep. [8]

PUBLISHED ANCHOR · EXACT

Per 10 mL plasma: 100 µL proteinase K stock (20 mg/mL); add 6.5 mL digestion buffer (5 M GITC, 25% Tween 20, 50 mM Tris pH 8, 25 mM EDTA), 56 °C ~1 h; add 33 mL binding buffer (3.5 M GITC, 45% isopropanol, 2.5% Tween 20, 10 mM Tris pH 8, 1 mM EDTA); add 400 µL of 2.5 mg/mL 400-nm silica beads and bind 10 min at room temperature.

PROPOSED SCREEN · STARTING POINTS

Proteinase K: 0.1, 0.2, 0.4 mg per mL original plasma
Digestion: 30, 60 min at 56 °C
Final GITC: 2.5–3.5 M; final isopropanol: 25–35%
Silica bead mass: 0.05–0.20 mg per mL original plasma
Bind: 5, 10, 20 min; elute 10 mM Tris pH 8.0 at 55–60 °C

These are deliberately bounded around the published UltraPrep environment; they are proposed DOE levels, not validated performance ranges.

Watch: irreversible adsorption, incomplete bead dispersion, residual guanidine/alcohol, bead over-drying and surface-area-dependent ultrashort loss. Boom's L6 formulation provides a mechanistic reference, not a drop-in cfDNA SOP. [7]

ROUTE B

PEG/NaCl carboxyl magnetic beads

Molecular crowding and salt drive reversible partitioning. Fragment cutoff shifts with PEG/salt and bead-stock ratio. [10]

PUBLISHED STOCK · EXACT

BOMB SPRI stock: 2.5 M NaCl, 20% (w/v) PEG8000, 10 mM Tris-HCl pH 8.0, 1 mM EDTA, 0.05% Tween 20 with 2% carboxyl bead slurry. It was developed for nucleic-acid cleanup—not as a validated direct-plasma cfDNA extraction. [9]

PROPOSED DIRECT-PLASMA ARM

First perform the same proteinase/SDS or proteinase/nonionic-detergent liberation block used across routes
Clarify lysate, then test 1.5×, 2.0×, 2.5×, 3.0× BOMB-like stock volume per lysate volume
Bind 10, 20, 30 min; test one versus two captures
Include 0.05% Tween 20 versus no Tween in the stock
Quantify both bead eluate and post-bind supernatant

This is an experimental adaptation. Plasma protein fouling and enzyme inhibition must be tested explicitly.

Watch: lower stock ratios impose size selection and may delete ultrashort molecules; higher PEG carryover can inhibit ligation. Carboxyl capture does not itself liberate nucleosomal/protein-bound DNA.

HIGH-VALUE ALTERNATIVE

Extraction-free / direct-to-library

Use a small, matched plasma aliquot to test whether solid-phase purification is the dominant loss. This arm is especially informative when both bead routes show high supernatant loss.

Evidence boundary

WO2020106893 reports a median 4.5× increase in molecules per µL versus an extraction process in one clinical-plasma comparison, plus more short fragments in platform-specific examples. Those data concern a microbial cfNA workflow and are patent evidence—not independent oncology validation. [19]

Gate

Advance only if short endogenous UMI families/mL rise without unacceptable inhibition, off-target background, VAF distortion, or loss of duplex confirmation.

PATENT-ONLY TUBE ARM

PEG preservative A

US20230365961 gives an exact stock embodiment: 33% (w/w) PEG—PEG8000 preferred—5% NaCl, 2% EDTA, 0.023% sodium azide, water; 1.5 mL added to ~8.5 mL blood yields nominal final concentrations near 4.95%, 0.75%, 0.30% and 0.00345%, respectively.

Test:

RUO collection prototype only; compare ±azide for downstream enzyme compatibility and benchmark against unmodified commercial tubes. This is patent evidence, not a validated current product. [21]

EXPLORATORY SURFACE

Lysine-modified zwitterionic beads

He et al. used 300 µL plasma, 2 mg MSP-ZEWB, glycine-HCl adsorption buffer plus 0.8% final PEG8000, 25 °C for 10 min, two 60% ethanol washes, and 40 µL 25 mM Tris pH 8.9 for 4 min.

Boundary:

Interesting adsorption arm, but the study included 24 colorectal-cancer patients, incomplete mutation comparison, and proprietary/patented bead manufacture. Treat as exploratory, not a lead until replicated. [22]

IV · MAKE HIDDEN MOLECULES VISIBLE

Ultrashort cfDNA changes the question.

A 166-bp peak is partly a property of the workflow. Multiple groups recovered a distinct ~50-nt/bp population when extraction and library chemistry were redesigned. [12][13]

Protein liberation

Crosslinked or nucleosome-associated DNA may require more proteolysis than naked spike-ins. Factor proteinase dose and time; never assume a synthetic ladder predicts endogenous recovery.

Diagnostic: endogenous 67-bp copies rise while naked spike recovery stays flat.

Single-strand visibility

Double-stranded library workflows undercount nicked, damaged and genuinely single-stranded molecules. Carry a ssDNA-compatible library arm through Stage 3.

Diagnostic: ssDNA library gains inserts <70 bp without a matching Qubit gain. [14]

Adsorption accounting

Low-input loss can occur on beads, tube walls and during transfers. Use low-bind versus standard polypropylene as a proposed factor and recover every discarded phase.

Diagnostic: mass balance across first eluate, second eluate and rebound supernatant.

Analytical benchmark

Hisano's PPIP used 0.5 mL plasma with NaCl, EDTA, SDS and proteinase K, followed by phenol/chloroform and isopropanol precipitation, revealing ~50-nt ssDNA. Use only as a manual benchmark under appropriate chemical controls—not a production workflow. [11]

LIVE MOLECULE MODEL

Budget the locus, not the tube.

Enter copies measured by a short endogenous ddPCR assay. The model propagates extraction, library conversion, VAF and finite read sampling.

copies / mL×plasma×R_ext×R_lib×VAF

Poisson and saturation calculations are planning aids, not an LoD claim. Clinical LoD requires a locked assay and replicate validation across specimens, lots, operators and interferents. [18]

Plasma input

Endogenous target

copies / mL

Extraction recovery

Library conversion

Variant allele fraction

Informative read pairs

reads

Target copies entering extraction9,600

Target copies in eluate6,240

Expected unique library molecules2,184

Expected mutant UMI templates (λ)4.4

P(≥1 mutant template)98.7%

P(≥3 mutant templates)81.1%

Expected unique templates observed2,128

Library sampled at least once97.4%

V · IMPLEMENTABLE MULTI-STAGE DOE

Screen broadly. Confirm on endogenous molecules.

Randomize within donor, block by donor and tube lot, and preserve matched aliquots for all finalists. Thresholds marked “proposed” are study gates—not published standards.

QUALIFY THE MATERIAL

Partition & baseline

Six or more donors for discovery; paired cancer plasma reserved for confirmation. Record recovered plasma volume, hemolysis, 67/180/300-bp copies/mL and fragment trace before pooling.

Gate 0 · proposed: exclude visibly cellular aliquots; predefine HMW and hemolysis limits before unblinding.

TUBE × TIME

Protect endogenous signal

Paired EDTA immediate/2 h/4 h versus Streck and PAXgene at 0/24/72 h within each current IFU; ≥6 donors, randomized processing order.

Gate 1 · proposed: 67-bp copies/mL within ±20% of time zero, long/short ratio ≤1.5× baseline, within-donor CV ≤15%.

DEFINITIVE SCREEN

Liberation × capture

Run a fractional factorial or definitive screening design across proteinase dose/time, detergent, silica versus carboxyl route, binding environment, bead mass/ratio, binding time and low-bind plastics. Include a commercial reference method.

Gate 2 · proposed: ≥1.25× short endogenous copies/mL versus reference with no rise in ≥300-bp contamination; inhibition ΔCq <0.5.

RESPONSE SURFACE

Optimize the winning neighborhood

For each viable surface, use Box–Behnken or central composite design around the best continuous factors. Run ≥3 donor pools and ≥3 process replicates per center point.

Gate 3 · proposed: predicted optimum confirmed within 15%; eluate + supernatant mass balance explains ≥80% of locus copies relative to the operational reference.

LIBRARY TRANSFER

Measure conversion, not reads

Cross 2–3 extraction finalists with dsDNA-UMI and ssDNA-compatible library arms at low and moderate input. Sequence all libraries to the same informative read-pair count.

Gate 4 · proposed: ≥1.25× duplex-UMI families/mL, noninferior on-target rate, stable VAF, and no new GC/insert-size distortion.

ONCOLOGY CONFIRMATION

Preserve mutant molecules

Test paired aliquots from 12–20 oncology specimens spanning low/high cfDNA and low/high VAF. Include matched leukocytes, contrived dilution series and relevant interferents.

Gate 5: lock acceptance criteria before the run; estimate recovery and detection with confidence intervals, then execute formal assay validation. [20]

Every extraction well67-bp + 180-bp endogenous ddPCRsynthetic 50/80/167/320-bp ladderinhibition control

Every discarded phasefirst-bind supernatantsecond elutionsecond-pass capture

Every libraryUMI families / mLduplex families / mLfamily-size distributionsaturation at fixed depth

Every conditionplasma input and eluate volumeoperator / lot / positionhands-on time and failure mode

STAGE BINDING

R_bind ≈ 1 − N_flow-through / N_input

STAGE RELEASE

R_release = (N_E1 + N_E2) / (N_input − N_flow-through)

Define N_input as endogenous locus copies entering the capture step after liberation; assay matched-volume flow-through, first eluate (E1) and second eluate (E2). Use the approximation only when sampling-volume and assay-dilution corrections have been applied.

EVIDENCE LEDGER

What anchors each experimental choice.

Primary papers, current official instructions and patents are separated by evidence type. Patent examples generate hypotheses; they do not establish current commercial composition or clinical performance.

NCI cfDNA Biospecimen Evidence-Based PracticeOfficial collection, processing, storage and extraction guidance.

Streck Cell-Free DNA BCT IVD IFUCurrent March 2026 document; disclosed tube claims and operating limits.

Streck patent WO2013123030IU/DU, EDTA and glycine embodiments; patent evidence is not a current product recipe.

PAXgene Blood ccfDNA Tube RUO IFU1.5 mL additive, 10 mL draw, eight inversions and current storage instructions.

PAXgene Blood ccfDNA Tube SDSCommercial formulation is proprietary; reportable hazardous ingredients are not enumerated.

PAXgene-family patent WO2017085321Aldehyde-free experimental embodiments using caspase inhibitor, PEGs, EDTA and DMPA.

06b

QIAGEN patent CN106132200B, Example 16Exact Alpha-tube component amounts plus internally inconsistent PEG300 concentration statements.

Boom et al., J Clin Microbiol (1990)Foundational GuSCN–silica extraction and L6 chemistry.

Raymond et al., PLOS ONE (2020)Open, direct-plasma silica-bead cfDNA protocol with exact reagent compositions.

Oberacker et al., PLOS Biology (2019)Open BOMB magnetic-bead platform and SPRI stock formulation.

DeAngelis et al., NAR (1995)Original solid-phase reversible immobilization on carboxyl beads.

Hisano et al., BMC Biology (2021)Proteinase/SDS/organic extraction benchmark exposing ~50-nt ssDNA.

Hudecova et al., Genome Research (2022)High-affinity bead plus ssDNA library workflow recovers a ~50-bp population.

Cheng et al., iScience (2022)Independent ultrashort single-stranded plasma cfDNA observation.

Burnham et al., Scientific Reports (2016)Single-stranded library preparation reveals ultrashort cfDNA.

Markus et al., Scientific Reports (2018)Short/long endogenous ddPCR, extraction recovery and library diversity.

Devonshire et al., Anal Bioanal Chem (2014)Recovery controls and strong fragment-size-dependent extraction bias.

Newman et al., Nature Biotechnology (2016)Molecule-aware error suppression and UMI-family concepts for ctDNA.

Godsey et al., Clinical Chemistry (2020)BLOODPAC ctDNA NGS analytical-validation framework.

Direct-to-library patent WO2020106893Extraction-free plasma library examples; patent/platform evidence, not oncology validation.

Razavi et al., Nature Medicine (2019)Matched leukocytes identify clonal-hematopoiesis and germline sources.

RUO preservative patent US20230365961PEG8000/NaCl/EDTA/azide stock embodiment; patent evidence only.

He et al., Frontiers in Oncology (2024)Exploratory lysine-modified zwitterionic bead extraction in colorectal-cancer plasma.