{"entity":{"id":"finite-molecule-sampling","kind":"term","name":"Finite-molecule sampling in blood tests","aka":["ctDNA molecular sampling","independent molecules versus sequencing reads"],"tldr":"A small blood sample contains a finite number of original DNA molecules. Sequencing copies of those molecules more deeply cannot recover a tumor molecule that never entered the analyzed sample.","summary":"Original molecule availability, recovery and technical errors place different limits on cell-free DNA analysis. A primary error-suppression study combined molecular barcoding and background correction to improve detection in a defined assay. Another primary comparison examined tumor-informed and tumor-naive assays on matched breast cancer samples. Their findings do not create one universal detection limit: sample input, tracked features and the complete workflow remain part of the result.","asOf":"2026-11-10","links":[{"label":"Molecular recovery and digital error suppression (2016)","url":"https://doi.org/10.1038/nbt.3520"},{"label":"Matched comparison of tumor-DNA assays (2023)","url":"https://doi.org/10.15252/emmm.202216505"},{"label":"FDA: statistical guidance for diagnostic-test studies (March 2007)","url":"https://www.fda.gov/media/71147/download"},{"label":"FDA: ctDNA for curative-intent solid tumour drug development (November 2024)","url":"https://www.fda.gov/media/183874/download"}],"tags":[],"related":["ctdna-shedding-negative-results","detection-versus-quantification","tumor-informed-assay-design"],"cancers":[],"sections":[],"technologies":["liquid-biopsy"],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":["For independent random target-molecule counts with Poisson mean lambda, the chance of none is exp(-lambda). At lambda = 1 this is 36.8%, so an average of one does not guarantee a molecule. In a fictional model, an evaluated aliquot supplies 4,500 independent amplifiable opportunities at one locus, each with a target-allele probability of 1 in 10,000. The expected target count is 0.45. If target counts follow a Poisson distribution, the chance of no target is exp(-0.45), 63.8%. Twenty independent loci with the same opportunity count and probability give an expected total of 9 and exp(-9), 0.012%, for none. Independence, equal recovery and one retained target being sufficient are assumptions of this model. These are not measured clinical false-negative rates. A tumour-DNA fraction cannot simply replace the target-allele probability without assumptions about copy number, variant multiplicity and which tumour clones carry the variant. More reads cannot recover an original molecule absent from the aliquot.","A worked example: in a multiplex PCR assay with a theoretical limit of 1 / (4,500 copies x variants), one sample not detected with 4,500 amplifiable copies was detected at 7.9 parts per million when re-assayed with 27,000 copies, and the authors note that detection at the theoretical limit rests on a single mutant molecule and is subject to the stochastic nature of sampling (Santonja 2023). A nominal detection limit from high-input validation material may not apply to a low-input clinical sample, so a laboratory with a 95% detection claim at 30 units of input cannot attach the same claim to a sample with 5 units without supporting data.","Analytical detection and clinical sensitivity answer different questions. FDA diagnostic-test guidance dated March 2007 defines clinical sensitivity by the proportion testing positive among people who have the specified condition, judged against the reference standard in the intended-use population. In a residual-disease study, disease definition, follow-up, draw schedule, evaluable samples and recurrence sites shape that denominator. An analytical limit measured in prepared material does not supply that clinical proportion, and a negative test does not establish absence of cancer.","Following a blood test from body to result separates biological availability, inclusion in the evaluated aliquot, recovery through processing, and the final calling rule. Low shedding, handling losses, background or a failed quality criterion can interrupt different steps. Repeated draws observe different times, but tumour biology and handling can make their errors correlated; they are not automatically independent repeats. A failed sample or an unbuildable reference needs its stated failure disposition rather than being silently called negative.","A validation report needs the evaluated plasma volume, recovered DNA and independent molecular coverage, together with the feature-level and combined calling rules. FDA's November 2024 drug-development guidance asks for representative clinical specimens, relevant input ranges, precision and accuracy, and comparability of any contrived reference samples. Testing one prepared input cannot show that every clinical sample has the same recovery, detection probability or outcome-based sensitivity."],"category":"Biomarkers"},"route":"/terms/finite-molecule-sampling/","neighbours":{"term":[{"id":"detection-versus-quantification","kind":"term","name":"Detection and quantification limits","route":"/terms/detection-versus-quantification/"},{"id":"measurement-to-benefit-evidence","kind":"term","name":"Evidence from measurement to patient benefit","route":"/terms/measurement-to-benefit-evidence/"},{"id":"ctdna-shedding-negative-results","kind":"term","name":"Tumor DNA shedding and negative plasma results","route":"/terms/ctdna-shedding-negative-results/"},{"id":"tumor-informed-assay-design","kind":"term","name":"Tumor-informed and tumor-naive assay design","route":"/terms/tumor-informed-assay-design/"}],"technology":[{"id":"liquid-biopsy","kind":"technology","name":"Liquid biopsy (ctDNA)","route":"/technologies/liquid-biopsy/"}]}}