Low-Biomass Microbiome Study Design: Controls, Contamination, and Method Selection

Low-biomass samples—tissue biopsies, skin swabs, air filters, cleanroom surfaces, meconium, and breast milk—amplify a problem that exists in every sequencing run: contaminant DNA from reagents, lab surfaces, and the operator can overwhelm the true microbial signal. This article provides a practical framework for designing a low-biomass microbiome study, covering the controls to include at each processing stage, how to match sequencing methods to sample types, and what statistical decontamination can and cannot fix.

Key takeaways

• Contamination is the dominant signal in low-biomass samples. Without negative controls, 80% or more of reads can come from reagent and environmental DNA rather than the sample itself.
• Negative controls must travel the full workflow. Collection blanks, extraction blanks, and library blanks—each sequenced alongside biological samples—are the only reliable way to identify which taxa are contaminants in your specific study.
• Published "kitome" lists are not a substitute for internal controls. Across seven published contaminant lists, 351 of 429 flagged genera appeared in only one list; none appeared in all lists.
• Method choice changes what contamination looks like. 2bRAD-M can generate species-level profiles from 1 pg of DNA and tolerates high host background, while shotgun metagenomics requires more input but delivers functional information.
• Statistical decontamination complements, but does not replace, wet-lab controls. Tools like decontam perform best when paired with sequenced negative controls and quantitative DNA measurements.

Why Low-Biomass Means High Risk

When a stool sample yields 10 million microbial reads and a negative control yields 500, the contamination ratio is negligible. When a skin swab yields 5,000 reads and the same negative control yields 500, contamination accounts for 10% of the data. For tissue biopsies or meconium, where true microbial biomass can be orders of magnitude lower, contaminant reads can dominate the profile entirely.

Salter and colleagues demonstrated this in 2014 by serially diluting a Salmonella bongori culture and sequencing each dilution with 16S rRNA gene primers. As the input DNA dropped, the proportion of reads matching Salmonella declined while reads matching common reagent contaminants—Pseudomonas, Acinetobacter, Stenotrophomonas—rose to dominance. At the lowest dilution, the contaminant taxa constituted the majority of the profile, creating a completely false picture of the sample's microbial content.

The MBQC baseline study later confirmed that contamination was "a frequent cause of variation" across 15 laboratories, with negative controls revealing both exogenous DNA (from reagents and handling) and endogenous cross-contamination between samples on the same plate. The key insight: contamination is not a binary yes/no variable but a continuum, and its impact scales inversely with sample biomass.

Controls That Actually Work

Collection Blanks

A collection blank captures contamination introduced during sampling—from the swab itself, the air at the collection site, the preservative solution, or the operator's gloves. To prepare one, open a sterile collection device at the sampling location, expose it to air for the same duration as a real sample collection, handle it identically, and seal it. For liquid samples, fill a collection tube with molecular-grade water or the same buffer used for real samples, carry it through the entire field protocol, and freeze it alongside the study specimens.

Collection blanks are the most frequently omitted control. A 2025 systematic review of 243 insect microbiota studies found that two-thirds included no negative controls at all, and only 13.6% sequenced their controls and used them to filter contaminants—a rate that did not improve over a decade of publications.

Extraction Blanks

Run one extraction blank—molecular-grade water or elution buffer processed through the entire DNA extraction protocol—for every batch of samples (typically one per 11–23 samples, or at minimum one per extraction kit lot). The extraction blank reveals contaminants introduced by the kit itself: residual bacterial DNA from column manufacturing, cross-reacting polymerases, and laboratory-air particulates that enter open tubes during the lengthy extraction workflow.

A 2019 study that tracked contamination over five years in an ancient-DNA cleanroom found that contaminant profiles shifted by researcher, month, and season, and that commercial kits contained higher microbial diversity and more human-associated taxa than home-made silica-based extraction protocols. This temporal and kit-level variability is exactly why published contaminant lists cannot replace study-specific blanks.

Library Preparation Blanks and PCR No-Template Controls

Include a no-template control (NTC) in each PCR plate and carry it through library preparation and sequencing. The NTC detects contamination introduced during amplification setup and index assignment. Index-hopping on Illumina platforms can also transfer reads between samples on the same flow cell; a well-designed plate layout that separates low-biomass samples from high-biomass samples and from each other can reduce this risk.

Mock Communities as Positive Controls

A mock community—a defined mixture of microbial species at known concentrations—serves as a positive control that validates the entire workflow. If a mock community with 10 species at equimolar concentrations returns 18 species with highly skewed abundances, something went wrong during extraction, amplification, or classification. While mock communities do not perfectly emulate the complexity of real samples, they provide an essential floor check: if the pipeline cannot correctly recover a simple, known input, it cannot be trusted for an unknown low-biomass sample.

Methods for Low-Input Samples

Once controls are in place, the next design decision is choosing a sequencing method that can extract usable data from limited starting material. Three approaches dominate the current landscape for low-biomass work, each with distinct input requirements, resolution, and contamination tolerance. For a broader overview of the trade-offs across all microbial sequencing approaches, see this guide on how to choose microbial sequencing methods.

2bRAD-M: Species-Level Resolution from Picograms

2bRAD-M uses a type IIB restriction enzyme (BcgI) to digest genomic DNA and then amplifies short, uniform fragments flanking restriction sites. Because it targets conserved sequence motifs rather than variable regions of the 16S rRNA gene, it achieves species-level taxonomic resolution for bacteria, archaea, and fungi simultaneously—and it does so from as little as 1 picogram of total DNA.

This ultra-low input requirement makes 2bRAD-M particularly valuable for samples where DNA yield is the primary bottleneck: FFPE tissue curls, single mosquito midguts, neonatal meconium, and breast milk (where >90% of total DNA can be human). A 2025 study applied 2bRAD-M to maternal-infant sample pairs and found it matched whole metagenome sequencing in abundance correlation while successfully profiling samples that were below the detection limit of conventional 16S rRNA gene sequencing. For researchers considering this approach, CD Genomics offers a dedicated 2bRAD-M analysis service for microbiome samples.

Amplicon Sequencing with Enhanced Controls

Standard 16S rRNA gene amplicon sequencing remains the most accessible method for many labs, but in the low-biomass context it requires several modifications. Reduce PCR cycles to 25 or fewer—higher cycle numbers preferentially amplify contaminant DNA that is present at low, constant levels. Use mechanical bead-beating for cell lysis rather than enzymatic lysis alone, as bead-beating has been identified as a major determinant of community composition. Assign samples and controls to plate positions that are not adjacent, to minimize cross-contamination during amplification and index assignment.

For researchers who need broad community profiling with well-established analysis pipelines, microbial diversity analysis by 16S/18S/ITS amplicon sequencing provides a mature, cost-effective entry point—provided the control framework described above is rigorously applied. Proper microbiome sample preparation with validated protocols for the specific sample type is an essential prerequisite.

Shotgun Metagenomics: Functional Data, Higher Input Requirements

Shotgun metagenomic sequencing captures all DNA in a sample, providing functional gene content, metabolic pathway reconstruction, and strain-level resolution. The trade-off for low-biomass samples is substantial: shotgun library preparation typically requires nanograms of input DNA, and the resulting libraries contain high proportions of host or contaminant reads. Host DNA depletion—enzymatic, chemical, or antibody-based—can improve the microbial read fraction, but each depletion step introduces its own contamination risk. For projects that need both taxonomic and functional information from samples above the picogram threshold, metagenomic shotgun sequencing with integrated host-depletion and contamination controls is the appropriate choice.

Statistical Decontamination and Batch Design

What decontam Does—and Does Not—Do

The decontam R package uses two statistical patterns to identify contaminant sequences: the frequency method (contaminants are more abundant when total DNA is low) and the prevalence method (contaminants appear more often in negative controls than in biological samples). When Karstens and colleagues benchmarked decontam against other approaches using a mock community dilution series, the frequency method removed 70–90% of contaminants without deleting expected sequences—outperforming simple presence/absence filters, which erroneously removed over 20% of true taxa.

The critical dependency: the frequency method requires quantitative DNA concentration measurements (e.g., PicoGreen or Qubit) for every sample, and the prevalence method requires sequenced negative controls. Without these inputs, decontam cannot be applied reliably.

Batch Design as Contamination Control

Randomizing sample processing order across extraction and library-preparation batches is one of the most effective and least expensive contamination-control measures. When all samples from one treatment group are extracted on the same day and all samples from the control group on another day, batch effects are perfectly confounded with the biological variable of interest—making it impossible to determine whether an observed difference is biological or technical.

Additional batch-design principles: record extraction batch, PCR batch, sequencing run, plate position, and date for every sample; include at least one technical replicate (ideally an extraction replicate) to quantify within-batch variation; process samples from each experimental group across multiple batches when feasible; and sequence all negative controls to the same depth as the lowest-biomass biological sample. For multi-center studies, these principles intersect with broader standardization requirements; see the companion article on microbiome standardization in multi-center studies.

Minimum Checklist for a Defensible Low-Biomass Study

FAQ

How many negative controls do I really need?

A minimum of one collection blank per sampling event, one extraction blank per extraction batch, one PCR no-template control per amplification plate, and one library blank per library-preparation batch. A practical rule of thumb is one control for every four biological samples when working with ultra-low-biomass specimens.

Can I use a published contaminant list instead of running my own negative controls?

No. Published "kitome" lists show minimal overlap across laboratories, kits, and time periods. Taxa that are genuine skin or tissue residents in one study can be reagent contaminants in another. Only study-specific negative controls sequenced alongside your samples can identify which taxa are contaminants in your specific workflow.

My sample yielded very little DNA. Which method gives me the best chance of getting usable data?

2bRAD-M is currently the sequencing approach with the lowest documented input requirement—reliable species-level profiles have been generated from 1 picogram of total DNA, and the method tolerates high host-DNA backgrounds. If your sample type and research question are compatible with taxonomic profiling rather than functional metagenomics, 2bRAD-M is the strongest candidate for ultra-low-input samples.

Is statistical decontamination enough if I cannot include controls?

It can reduce the impact of contamination, but it cannot replace controls. The prevalence-based decontam method requires negative controls to function at all. The frequency-based method can operate without controls but needs quantitative DNA measurements for every sample, and it cannot distinguish between a low-abundance true resident and a low-abundance contaminant without the prevalence signal from negative controls. Wet-lab controls and computational decontamination are complementary, not interchangeable.

How do I know if my negative controls are clean enough?

There is no universal threshold for an acceptable level of contamination in negative controls. The practical standard is that contaminant reads in blanks should be substantially lower than reads in the lowest-biomass biological sample being analyzed, and the contaminant taxa should not be the primary drivers of diversity or differential-abundance results. Report the composition of your negative controls transparently, and describe your decontamination criteria so reviewers can assess adequacy.

References

1. Salter SJ, Cox MJ, Turek EM, et al. Reagent and laboratory contamination can critically impact sequence-based microbiome analyses. BMC Biology. 2014;12:87. doi:10.1186/s12915-014-0087-z
2. Eisenhofer R, Minich JJ, Marotz C, Cooper A, Knight R, Weyrich LS. Contamination in low microbial biomass microbiome studies: issues and recommendations. Trends in Microbiology. 2019;27(2):105-117. doi:10.1016/j.tim.2018.11.003
3. Davis NM, Proctor DM, Holmes SP, Relman DA, Callahan BJ. Simple statistical identification and removal of contaminant sequences in marker-gene and metagenomics data. Microbiome. 2018;6(1):226. doi:10.1186/s40168-018-0605-2
4. Karstens L, Asquith M, Davin S, et al. Controlling for contaminants in low-biomass 16S rRNA gene sequencing experiments. mSystems. 2019;4(4):e00290-19. doi:10.1128/mSystems.00290-19
5. Sinha R, Abu-Ali G, Vogtmann E, et al. Assessment of variation in microbial community amplicon sequencing by the Microbiome Quality Control (MBQC) project consortium. Nature Biotechnology. 2017;35(11):1077-1086. doi:10.1038/nbt.3981
6. Sun Z, Huang S, Zhu P, et al. Species-resolved sequencing of low-biomass or degraded microbiomes using 2bRAD-M. Genome Biology. 2022;23(1):36. doi:10.1186/s13059-021-02576-9
7. Hou S, Jiang Y, Zhang F, et al. Unveiling early-life microbial colonization profile through characterizing low-biomass maternal-infant microbiomes by 2bRAD-M. Frontiers in Microbiology. 2025;16:1521108. doi:10.3389/fmicb.2025.1521108
8. Agudelo J, Miller AW. Impact of study design, contamination, and data characteristics on results and interpretation of microbiome studies. mSystems. 2025;10(9):e00408-25. doi:10.1128/msystems.00408-25