henhao
Sign in
Molecular BiologybioRxiv
Preprint

Characterizing mitochondrial copy number variation and PCR amplification bias as sources of quantitative constraints in DNA metabarcoding

Lisa WolanyiD, Kevin Klinkenborg, Florian LeeseiD, Dominik BuchneriD

PDFDOICode
Post to X
Threads on XPost to BlueskyThreads on Bluesky
In-depth full-text digest available · Jump to digest ↓

Abstract

DNA metabarcoding is a central tool in biodiversity research and monitoring, producing detailed taxa lists with comparatively little time and effort. One of its limitations, however, is the lack of quantitative data on biomass or abundance. This limitation has two main reasons: 1) DNA template copy number variation and 2) primer-induced PCR amplification bias. Many metabarcoding markers are mitochondrial and mitochondrial DNA copy numbers vary in animal tissues, potentially decoupling sequence counts from biomass. Additionally, primer mismatches can lead to taxon-specific amplification biases, for which PCR cycle calibration has been proposed as a solution. To mechanistically study both effects, we constructed and analyzed mock communities of different arthropod species. We combined digital PCR and COI metabarcoding to quantify relationships between biomass, mitochondrial copy number, and metabarcoding reads. Mitochondrial DNA copy numbers per biomass varied strongly within and among the different taxa. Metabarcoding reads did not reflect input mitochondrial DNA copies without a correction. Attempts to correct for amplification bias via PCR cycle calibration failed as read proportions remained stable across cycles. We therefore mathematically derived an approach to estimate relative amplification bias and initial mitochondrial DNA copy numbers in a sample based on a non-exponential amplification bias model and demonstrate its applicability. Still, the detected high variation in mitochondrial copy numbers and derived prerequisites necessary to calculate amplification efficiencies and mitochondrial copy numbers limit the practical application. Our study highlights fundamental constraints of quantitative metabarcoding and underscores the need for additional methodological approaches for quantitative insights while delivering essential conceptual insights.

02 /Specification & Classification

Click any tag to browse matching papers
Citations
1
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (17 Aug 2026) · CC BY 4.0

COI metabarcoding of 81 arthropod mock communities shows PCR amplification bias is taxon-specific but non-exponential, so cycle calibration fails while a constant-bias model recovers mtDNA copies.

Metabarcoding reads are compositional and distorted by marker copy-number variation and primer-driven amplification bias, so they cannot directly stand in for biomass or abundance, and prior cycle-calibration corrections have mixed empirical support. Jointly measuring biomass, dPCR mtDNA copies and reads in mock communities separates the two biases and yields a non-exponential, reference-relative efficiency correction.

Key results

  1. ShownShares of biomass were only loosely matched by shares of mtDNA copies measured by digital PCR (dPCR). Copy number per unit biomass differed within and among taxa, and some orders were consistently skewed.

    Biomass and mtDNA copy ranks matched in 112 of 210 species-sample pairs (53.3 %); beyond a 1.5-fold band, Amphipoda was underrepresented in 78 % of pairs and Diptera overrepresented in 56.8 %; log2 fold-changes spanned 2.63 to −3.91. · Variation of biomass, mtDNA copy number, and…; Tables

  2. ShownWithout correction, read shares tracked mtDNA copy shares for taxa that amplify well but badly underrepresented taxa that amplify poorly. Errors added up across the path from biomass to copies to reads.

    fwh2, cycle 20: copy-read rank concordance 75.7 % (BF3: 74.8 %); Amphipoda under-represented in every sample and Hymenoptera in 90.2 % of pairs; Blattodea, Diptera, Lepidoptera within 1.5-fold in 63.6 %, 81.8 %, 72.5 %; copy-read fold-changes 2.98 to −12.70; biomass-read rank concordance 44.3 % (BF3: 43.8 %), fold-changes 4.20 to −15.82. · Variation of biomass, mtDNA copy number, and…; Discussion; Tables

  3. ShownPCR cycle calibration failed: read proportions did not shift systematically across 6 to 20 first-step cycles, so the cycle-based regression recovered relative copy proportions no better than raw reads did.

    R^2 against dPCR copy proportions (81 communities): −0.064 uncorrected vs −0.039 cycle-calibrated; median Aitchison distance (67 held-out communities): 3.78 uncorrected vs 3.51 cycle calibration. · PCR cycle calibration; mtDNA copy number calculation under different…; Figure legends

  4. ShownIn a derived two-stage model, mismatch bias acts only until primers replace the template binding sites (first two cycles), giving a constant relative efficiency. This model predicted held-out copy proportions better than the exponential, cycle-dependent model.

    Calibrated on 14 Blattodea two-species communities, tested on 67 others: R^2 0.740 (constant bias) vs 0.489 (cycle-dependent); median Aitchison distance 0.54 vs 1.55; constant-bias advantage held at cycles 6-16, was minor at 18, and predictions were near-identical at 20 by construction. · Mathematical framework for amplification bias; mtDNA copy number calculation under different…; Figure legends

  5. SuggestedWith the second primer pair (BF3), taxon efficiencies differed from those under fwh2, yet the constant-bias correction still improved copy prediction. The authors take this primer dependence as evidence that primer mismatch, not a primer-independent source, drives the bias.

    BF3 constant-bias relative efficiencies: Amphipoda Ei = 1.30, Diptera Ei = 0.97, Hymenoptera strongly underrepresented with both primers; R^2 rose from 0.43 (uncorrected) to 0.90 (corrected). · mtDNA copy number calculation under different…; Discussion

Methods, data and reproducibility

81 mock communities of five arthropod species were homogenized; mtDNA copies were quantified by species-specific dPCR and COI amplified with fwh2 primers at 6-20 first-step cycles (BF3 at 20). Reads were compared with biomass and copies via log2 fold-changes and rank concordance, then corrected by cycle calibration or a derived two-stage constant-bias model.

  • Arthropod mock communities (two- to five-species mixes, 5% to 90% biomass per species) of Acromyrmex sp., Blaptica dubia, Galleria mellonella, Gammarus sp., Hermetia illucens · arthropods (five species across five orders) · n = 81 communities (84 assembled, 3 excluded)
  • dPCR mtDNA copy measurements per species per community (extraction replicates A and B averaged; 1:1000 or 1:100 dilution) · arthropods (five species) · n = 81 communities
  • COI fwh2 amplicon reads, four extraction replicates (plates A-D) each at eight first-step cycle numbers (6 to 20), Illumina NovaSeq · arthropods (five species) · n = 1,271,494,502 raw reads; 1,088,663,970 retained · PRJEB110937 (ENA, demultiplexed raw reads)
  • COI BF3/BR2 amplicon reads, four extraction replicates at 20 first-step plus 25 second-step cycles, Illumina NovaSeq · arthropods (five species) · n = 741,085,198 raw reads; 473,220,495 retained · PRJEB110937 (ENA, demultiplexed raw reads)
  • Calibration subset: two-species communities pairing Blattodea (reference) with one other taxon; remaining communities held out for evaluation · arthropods · n = 14 calibration; 67 evaluation
  • Ethanol-only negative controls placed so every row and column of the 96-well plate held at least one · n = 12
Methods & tools
QuantStudio Absolute Q digital PCR (dPCR), Primer3 (species-specific dPCR primer/probe design), qPCR (dPCR assay specificity check), Two-step PCR COI metabarcoding (fwhF2/fwhR2n; BF3/BR2), Illumina NovaSeq, demultiplexer2 v 1.1.4, APSCALE pipeline v. 4.3.0, BOLDigger3 v. 3.0.2, Hamilton Microlab Vantage liquid handler, Wilcoxon signed-rank test, PCR cycle calibration (Shelton et al.), Aitchison distance on CLR-transformed proportions
Code
https://github.com/DominikBuchner/quantitative_metabarcodinghttps://github.com/DominikBuchner/demultiplexer2
Software & environment
demultiplexer2 v 1.1.4; APSCALE v. 4.3.0 (default settings); BOLDigger3 v. 3.0.2; full analysis script on GitHub and as supporting information 10
Reproducibility gaps
Extraction protocol cited as on protocols.io without a link in the text; dPCR primer/probe sequences, community design and per-sample tables are in supporting information; some BF3 fold-change results are given as data not shown.
Limitations (6)
  • Because mtDNA copies per unit biomass vary so widely, even exact copy estimates would not give reliable biomass or abundance; for eDNA, differences in shedding rates across taxa and life stages add further error. · stated by the authors
  • The correction only works if every taxon in a sample is known, along with its efficiency relative to a co-occurring reference taxon. Each efficiency needs its own two-species mock community, which is unrealistic for Malaise trap or eDNA samples. · stated by the authors
  • The 1.5-fold tolerance is arbitrary. Because compositional data are bounded at 100%, fold-changes are asymmetric: underrepresentation can grow without limit while overrepresentation is capped. · stated by the authors
  • Whether the constant-bias model transfers to other mitochondrial markers (cytochrome b, 16S or 12S rDNA) has not been tested. · stated by the authors
  • Efficiencies come from five species, each from a single source, with one reference taxon (Blattodea) in one lab workflow. Intraspecific variation at primer sites and the complexity of natural samples were not assessed. · noted in this summary
  • Species with zero reads were excluded from the fold-change calculations, and samples with zero reads were left out of Figure 4, which may understate the bias for taxa that amplify very poorly. · noted in this summary

Who should read:Read this if you use COI or other mitochondrial metabarcoding to infer biomass or abundance and want mechanistic evidence on why PCR cycle calibration fails and how primer replacement limits amplification bias.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISMoinard '23Gold '23Shelton '22Braukmann '18Yang '21Kamenova '25
81% semantic matchEcology
S. Moinard, D. Piau, F. Laporte+4 more·bioRxiv·Oct 3, 2023