Summary

Antimicrobial resistance genomic surveillance is moving beyond the simple detection of resistance determinants toward a more contextual view of how resistance emerges, spreads, and persists across microbial populations. The 2026–2030 National Antimicrobial Resistance Monitoring System (NARMS) Strategic Plan reinforces whole-genome sequencing, predictive analytics, plasmid tracking, metagenomics, One Health data integration, and exploration of advanced technologies. Isolate WGS, metagenomics, long-read sequencing, cell-resolved microbial genomics, and microbial single-cell transcriptomics each resolve different layers of AMR biology.

Research-use note: The surveillance, resistance, and microbial genomic examples discussed below describe research and method-development applications. The services described are not intended for clinical diagnosis or individual treatment decisions.

Key Takeaways

• NARMS 2026–2030 explicitly prioritizes WGS, predictive AMR analytics, plasmid tracking, metagenomics, One Health data integration, and exploration of advanced technologies.
• NARMS does not specifically recommend single-cell microbial sequencing; this article evaluates it as one complementary research direction.
• Isolate WGS, short-read metagenomics, long-read metagenomics, single-microbe genomics, and microbial single-cell transcriptomics answer different questions.
• Detecting an antimicrobial resistance gene does not always resolve its microbial host or mobile-element context.
• Host–plasmid association is not the same as direct proof that horizontal gene transfer occurred.
• Resistance, tolerance, and persistence are different biological concepts and require different evidence.
• Study-specific cell counts, genome recovery rates, and performance values should not be treated as universal service specifications.

Figure 1. Antimicrobial resistance genomic surveillance is expanding from resistance detection toward genomic context and microbial population resolution.

NARMS 2026–2030: From Resistance Trends to Mechanism-Aware Surveillance

NARMS was established in 1996 as a collaboration among the U.S. Food and Drug Administration, Centers for Disease Control and Prevention, U.S. Department of Agriculture, and state and local partners. Over three decades, the program evolved from antimicrobial susceptibility testing of human Salmonella isolates into an integrated One Health surveillance system that also uses whole-genome sequencing. The 2026–2030 plan centers on strengthening surveillance, expanding collaborative networks, and embracing innovation. Its Goal 2 is especially relevant to antimicrobial resistance genomic surveillance.

The plan calls for fuller use of WGS, predictive AMR analytics, plasmid tracking tools, and other bioinformatics processes. It also calls for evaluating the best uses of metagenomics, improving integration of One Health data, and exploring advanced technologies that can support discovery, gene tracking, predictive models, and antimicrobial stewardship. The scale of the existing program is important context. For 2021–2025, NARMS reported AMR prevalence data from more than 430,000 human clinical isolates, 210,000 food-producing animal isolates, and enteric bacteria from more than 170,000 retail meat samples. These are program counts, not assay-performance benchmarks.

What Goal 2 Actually Says

The official plan identifies several priorities:

• Objective 2.1: fully harness WGS, predictive analytics, plasmid tracking, and bioinformatics;
• Objective 2.2: evaluate how metagenomics can augment AMR monitoring;
• Objective 2.3: improve One Health data interoperability and interpretation;
• Objective 2.4: explore advanced technologies to support discovery, prediction, gene tracking, learning, and data sharing.

The plan also notes that culture-independent diagnostics can reduce bacterial culture availability, creating a need for more flexible surveillance methods.

What the NARMS Plan Does Not Say

The plan does not name single-cell microbial genomics as a required NARMS method and should not be presented as an official endorsement of any specific commercial single-cell platform. In this article, single-microbe genomics and microbial single-cell transcriptomics are therefore evaluated as complementary research approaches that may help answer questions left unresolved by existing workflows. For the official program language, see the [1].

What Current AMR Surveillance Can—and Cannot—Tell Us

No single AMR method provides phenotype, genomic context, host linkage, mobile-element structure, and cell-state information at once. A better way to compare surveillance technologies is to ask what each one measures most directly.

Figure 2. AMR surveillance methods answer different questions, from phenotype and isolate genomics to host context and microbial cell states.

Recent work illustrates why these methods should not be arranged as a replacement ladder. A 2024 BMC Genomics study showed that metagenomic assemblies often break around ARGs, making taxonomic origin and genomic context harder to resolve. [3]. Long-read metagenomics can recover information that short-read workflows miss. A 2025 Frontiers in Microbiology study used nanopore sequencing, methylation signatures, and strain haplotyping to improve plasmid-host and strain-level AMR interpretation in chicken fecal samples. [4]. The practical lesson is simple: Use the method that resolves the remaining uncertainty.

If isolate WGS already answers the question, a cell-resolved layer may add little value. If long-read metagenomics resolves plasmid context and host linkage, it may be the more direct design. If host assignment remains ambiguous, single-microbe genomics becomes more relevant. Researchers who need genome-level analysis from individual cells can also review our Single-cell Whole Genome Sequencing Service.

Challenge 1 — Who Carries the Resistance Gene?

One of the most important questions in AMR research is not merely whether an antimicrobial resistance gene is present. It is: Which organism, strain, or cell carries it?

Culture-Based WGS Has a Sampling Boundary

Isolate WGS can provide excellent genomic context once a bacterium has been recovered and cultured. The limitation is sampling. Many host-associated and environmental microorganisms remain uncultured or difficult to recover under routine laboratory conditions. Low-abundance taxa may also be missed when culture conditions favor faster-growing organisms. This does not make culture-based WGS obsolete. It means isolate-based surveillance represents organisms successfully recovered under the chosen conditions, which matters in complex communities such as wastewater, gut, animal, and food-associated microbiomes.

Metagenomic ARG Detection Is Not Always Host Assignment

Metagenomics avoids the need to isolate every organism before sequencing. However, detecting an ARG in a mixed community does not guarantee that its host can be assigned with high confidence. The problem is especially difficult when:

• the ARG occurs in multiple species;
• the surrounding genomic sequence is repetitive or mobile;
• several closely related strains coexist;
• the ARG lies on a plasmid;
• coverage is low;
• the relevant contig is short or fragmented.

The 2024 assembly study noted above found that common metagenomic assemblers could identify the ARG repertoire but often failed to recover the full diversity of genomic contexts in complex samples. Long reads can improve this situation. They may span larger genomic regions, support strain phasing, and provide additional host-linking signals. But long reads do not guarantee complete host assignment in every community.

When Single-Microbe Genomics Adds Value

Single-microbe genomics approaches the problem from another direction. Instead of reconstructing all relationships from a mixed pool of DNA, individual microbial cells are separated before their genomes are amplified and sequenced. The resulting single amplified genomes, or SAGs, can provide a physical within-cell association between the recovered microbial genome and recovered ARG or mobile-element sequences. That statement needs an important qualifier: The association is only as complete as the genomic material recovered from that cell. Whole-genome amplification may be uneven, and some regions may be missing. A gene not detected in a SAG is therefore not necessarily absent from the original cell. Interpretation should consider:

• SAG completeness;
• coverage uniformity;
• contamination;
• taxonomic consistency;
• ARG alignment quality;
• mobile-element reconstruction quality;
• evidence from additional cells or orthogonal methods.

For complex microbial communities where ARG host assignment remains unresolved, Microbial Single-Cell Sequencing Services can be considered as a research option.

Challenge 2 — Which Plasmid Is in Which Host, and Has HGT Occurred?

Resistance can spread through bacterial clones, but resistance genes can also move through mobile genetic elements. Plasmids are particularly important because they can carry resistance determinants across strains, species, hosts, and environmental compartments. A 2026 review in JAC-Antimicrobial Resistance summarizes genomic evidence for plasmid-mediated AMR across food, animals, humans, and environmental reservoirs. The review reinforces why One Health surveillance increasingly needs to consider the mobile element as well as the bacterial lineage. [9].

Plasmid Detection Is Not the Same as Host Assignment

Finding a plasmid-associated resistance gene in a metagenome is informative. But several distinct questions remain:

• Is the plasmid fully reconstructed?
• Which microbial host carries it?
• Is the association present in one cell or inferred computationally?
• Is the same plasmid present in multiple hosts?
• Are similar plasmids being mistaken for one transmission event?

Isolate long-read sequencing can often provide strong plasmid structure and host information for cultured organisms. Long-read metagenomics can also improve plasmid reconstruction and, in some settings, use methylation or other signals to support host assignment. Single-microbe genomics offers another route because plasmid and chromosomal DNA recovered from the same microbial cell can be analyzed together.

Host–Plasmid Association Is Not Direct Observation of Transfer

This is a critical interpretation boundary. If similar plasmid sequences are found in different bacterial hosts, the data may support a candidate horizontal gene transfer network. That does not mean the experiment directly watched a transfer event occur. A strong HGT claim may require additional evidence such as:

• longitudinal sampling;
• near-identical mobile-element sequences across distinct hosts;
• temporal ordering;
• conjugation experiments;
• strain-resolved epidemiological evidence;
• independent plasmid reconstruction;
• compatible ecological or exposure metadata.

Therefore, the correct interpretation is usually: The data support host–mobile-element associations and may support reconstruction of candidate HGT relationships. It is not: The data prove that transfer occurred at a specific time and place. This distinction protects the analysis from converting a genomic association into a causal event.

Figure 3. Three AMR surveillance gaps involve assigning resistance genes to hosts, resolving mobile-element context, and detecting rare microbial states.

Challenge 3 — Resistance Is Not Only a Gene-Presence Problem

Resistance-gene carriage is only one layer of microbial behavior. Cells with similar genomes can occupy different physiological states under antimicrobial stress. That is why resistance, tolerance, and persistence should not be treated as interchangeable terms.

Resistance, Tolerance, and Persistence Are Different

Antimicrobial resistance generally refers to the ability of a microorganism to grow at antimicrobial concentrations that inhibit a susceptible comparator. The phenotype may be associated with heritable resistance genes, mutations, or other stable mechanisms. Tolerance refers to the ability of a population to survive antimicrobial exposure for longer without necessarily showing an increased MIC. Persistence describes a phenotypic state in which a subpopulation survives an antimicrobial exposure that kills most genetically similar cells. The surviving state is not equivalent to a stable resistance genotype. These distinctions matter because each question calls for different evidence.

Genomics Shows Potential; Transcriptomics Shows State

A genome can show whether a cell carries a resistance determinant. A transcriptome can show which genes are being expressed under a defined condition. Cell-resolved transcriptomics may therefore help reveal microbial subpopulations that activate:

• efflux-associated programs;
• DNA damage responses;
• oxidative-stress responses;
• metabolic adaptations;
• growth-state changes;
• other condition-specific stress programs.

This can complement genomic carriage data. It does not replace phenotype. A 2025 Journal of Hazardous Materials study used single-cell RNA sequencing in a soil-isolated Escherichia coli experimental system under ciprofloxacin exposure. The authors used cell-resolved transcriptional data to describe heterogeneous subgroups during resistance evolution. The study is useful as a proof of research capability, but it was a defined laboratory evolution model and should not be generalized as a universal AMR surveillance performance result. [8].

Transcriptomic State Does Not Identify a Persister by Itself

A transcriptional cluster that appears dormant or stress-adapted is not automatically a persister population. Persister identification requires phenotypic evidence related to survival under antimicrobial exposure and recovery after treatment. Useful validation may include:

• antibiotic survival assays;
• time-kill measurements;
• regrowth after drug removal;
• matched phenotype and transcriptome experiments;
• targeted validation of state-associated pathways.

Microbial single-cell RNA analysis is most useful when the research question concerns heterogeneous cell states, not when a transcriptome is being used as a substitute for susceptibility testing. Researchers with existing single-cell RNA data can also review our Single-Cell RNA-Seq Data Analysis Service.

How Cell-Resolved Microbial Omics Can Complement AMR Surveillance

Cell-resolved microbial omics is most useful when the unresolved question depends on the identity or state of individual cells. It should be treated as a complement to established surveillance technologies.

Figure 4. Cell-resolved microbial genomics and transcriptomics provide complementary evidence for AMR host context and microbial heterogeneity.

Track 1 — Single-Microbe Genomics

A conceptual genomic workflow includes: Complex microbial sample → individual-cell isolation → cell lysis and genome amplification → single amplified genomes → taxonomic and strain analysis → ARG and mobile-element annotation → host–element association Depending on project design and data quality, outputs may include:

• SAG assemblies;
• taxonomic assignments;
• strain comparisons;
• ARG annotations;
• plasmid or phage fragments;
• candidate host–mobile-element associations;
• comparative genomics.

The strongest use case is not “sequence everything at single-cell resolution.” It is a specific question that benefits from individual-cell genomic context.

Track 2 — Microbial Single-Cell Transcriptomics

A transcriptomic workflow addresses a different problem: Defined biological condition → individual microbial cells → cell-resolved transcript capture → expression matrix → transcriptional clustering → state-associated pathway analysis This track may be useful when researchers need to ask:

• Which cells activate a stress response?
• Are several transcriptional states present in one strain?
• Does antibiotic exposure change the proportion of specific states?
• Which pathways distinguish responsive and nonresponsive subgroups?

Again, these are transcriptional questions.

Which Question Maps to Which Data Layer?

For broader cell-resolved study design, see our Single-cell Sequencing Service.

Project-Planning Checkpoint

Before adding a new modality, ask:

1. What relationship is unresolved in the current data?
2. Can a better isolate-based or long-read design answer it?
3. Is cell-level physical association essential?
4. Is the question genomic, transcriptional, or phenotypic?
5. What independent validation will be required?

The most efficient project adds the minimum evidence layer needed to answer the unresolved question.

Recent Published Evidence: What Single-Cell Methods Have Actually Demonstrated

Published studies now provide concrete examples of where cell-resolved microbial sequencing can add information. Their sample types and experimental settings must remain explicit; study-specific results are not general service specifications.

Activated Sludge: Linking ARGs and Mobile Elements to Microbial Hosts

Zhang and colleagues applied high-throughput single-cell sequencing to an activated-sludge microbiome. The study reported 15,110 sequenced single cells and 2,454 SAG bins. It detected 1,137 antibiotic resistance genes, 10,450 plasmid fragments, and 1,343 phage contigs. These are study-specific observations, not universal specifications. Their main relevance is the ability to examine resistance genes and mobile elements in the context of individual microbial genomes. The paper was published in Environmental Science and Ecotechnology. [6].

Antibiotic-Treated Human Gut Microbiome: A Single-Patient Research Example

Ye and colleagues used single-cell sequencing to study the gut microbiome of a hospital patient receiving antimicrobial treatment. The study examined population-level bacterial adaptation and ARG distribution across recovered microbial genomes. This is a single-patient observational research example. It supports feasibility in a complex human gut sample but should not be generalized to population-level surveillance accuracy, clinical outcome prediction, or diagnostic performance. The paper appeared in mSystems. [7].

Bacterial Transcriptional Heterogeneity Under Ciprofloxacin Exposure

Wang and colleagues used single-cell RNA sequencing to examine heterogeneous bacterial states during an experimental ciprofloxacin resistance-evolution model. The research supports the idea that bacterial populations can contain transcriptionally distinct subgroups under antimicrobial pressure. But the study used a soil-derived E. coli isolate in a defined experimental setting. The identified transcriptional subgroups should not be treated as universal markers for resistance, tolerance, or persistence across species or surveillance environments. This is a useful example of state-resolution, not a replacement for AST.

Comparator Evidence: Long-Read Metagenomics Can Also Improve Host Context

A balanced design should also consider methods that may answer the same question without single-cell sequencing. Bloemen and colleagues showed that long-read nanopore metagenomics with methylation analysis and strain haplotyping could:

• link plasmids to bacterial carriers;
• resolve strain-level genomic variation;
• identify resistance-associated mutations;
• improve interpretation of fluoroquinolone resistance in complex fecal samples.

This shows that host-context problems are not exclusive to single-cell solutions. Method choice depends on sample complexity, coverage, required resolution, and whether within-cell association is necessary.

Practical Study-Design Scenarios

Scenario 1 — Wastewater Resistome With Uncertain Hosts

A wastewater metagenomics project identifies several ARGs of interest. Their abundance is clear, but short-read assembly does not resolve which taxa carry them. A practical decision path is:

1. Review read mapping and assembly context around the ARGs.
2. Determine whether long-read metagenomics is likely to improve continuity and host assignment.
3. If host ambiguity remains and the biological question depends on cell-level linkage, consider single-microbe genomics.
4. Validate high-priority ARG–host relationships using an independent method when feasible.

This avoids using single-cell sequencing simply because it offers higher resolution.

Scenario 2 — One Health Plasmid Tracking Across Animal, Food, and Environmental Samples

A similar plasmid-associated ARG appears in animal, food, and wastewater samples. The immediate temptation is to infer transmission. That would be premature. A stronger design would combine:

• isolate WGS where isolates are available;
• long-read plasmid reconstruction;
• host assignment;
• strain phylogeny;
• sample metadata;
• temporal and geographic context.

If important parts of the microbial community remain uncultured or host assignment remains ambiguous, single-microbe genomics can add another evidence layer. Even then: Finding the same or similar plasmid in several hosts does not by itself prove a transmission route.

Scenario 3 — Antibiotic-Exposed Population With Hidden Cell States

A laboratory bacterial population shows little change in MIC, but bulk RNA-seq suggests a strong stress response. The research question is whether a minority subpopulation is driving that expression pattern. Microbial single-cell transcriptomics may help separate:

• high-stress and low-stress states;
• growth-associated and slow-growth states;
• subgroups with different pathway activation.

If the goal is specifically to identify persisters, the study must also include phenotypic survival and recovery evidence. This is a good example of how the final biological claim determines the required validation.

Common Failure Modes and Troubleshooting

Single-cell microbial workflows introduce biases that must be considered alongside sequencing output.

Cell Isolation and Lysis Bias

Different microbial taxa can differ greatly in:

• cell-wall structure;
• size;
• aggregation behavior;
• membrane properties;
• sensitivity to lysis conditions.

A protocol that works well for one organism may under-represent another. Practical check: review expected taxa, cell morphology, debris, aggregation, and lysis compatibility before interpreting low recovery as biological absence. For mixed communities, a mock community or other known reference can help reveal taxon-specific recovery bias.

Incomplete or Uneven SAG Recovery

Whole-genome amplification can be highly uneven. Some genomic regions may be amplified strongly while others are poorly recovered. This can affect:

• assembly completeness;
• variant detection;
• ARG recovery;
• plasmid reconstruction;
• strain comparison.

Therefore: ARG absence from a SAG does not prove biological absence from the original cell. Negative findings should be interpreted in the context of genome completeness and coverage.

Cross-Cell Contamination or Mixed Genomes

Contamination can create false host assignments. A resistance gene from one cell can be incorrectly associated with another genome if mixed material enters the same reaction or if downstream processing introduces contamination. Useful checks include:

• genome contamination metrics;
• taxonomic consistency;
• conflicting marker genes;
• unexpected multi-lineage signatures;
• recurrence across independent cells;
• orthogonal confirmation for high-priority links.

Plasmid Recovery Is Not Guaranteed

Plasmids can be difficult to reconstruct. Challenges include:

• repetitive regions;
• copy-number variation;
• amplification bias;
• incomplete assembly;
• similarity among related plasmids.

A plasmid-associated fragment detected in one SAG is not necessarily a complete plasmid genome. Where plasmid structure is central, long-read isolate sequencing, long-read metagenomics, or targeted validation may be needed.

Rare-Taxon Detection Is Sampling-Dependent

A method that can detect rare cells is still limited by how many cells were actually sampled. Failure to observe a rare population can reflect:

• insufficient sampled cells;
• low abundance;
• cell loss during preparation;
• isolation bias;
• lysis bias;
• sequence recovery failure.

“Not detected” should not be converted into “not present.”

Microbial RNA Is Technically Difficult

Microbial transcriptomics has different constraints from mammalian scRNA-seq. Bacterial cells contain relatively little RNA. Ribosomal RNA can dominate. Transcription can also change rapidly during sampling and handling. Potential artifacts include:

• stress responses induced during processing;
• state shifts caused by delayed stabilization;
• species-dependent permeabilization bias;
• low transcript recovery;
• high technical sparsity.

Experimental handling is therefore part of the biological interpretation.

Computational Integration Can Overstate Evidence

Integrated datasets often combine ARG annotation, taxonomy, plasmid sequence, strain identity, and transcriptional state. These layers do not all have the same evidence strength. Reports should distinguish:

• direct within-cell observation;
• assembly-supported association;
• database-based annotation;
• computational inference;
• biological hypothesis.

If a relationship depends on several inferential steps, the conclusion should reflect that uncertainty.

What Must Be Validated Before Calling a Resistance Mechanism

Cell-resolved microbial omics can provide strong evidence for association and prioritization. It can support:

• ARG–host associations;
• plasmid–host associations;
• candidate strain relationships;
• cell-state heterogeneity;
• resistance-associated transcriptional programs;
• hypotheses about mobile-element spread.

By itself, it generally cannot establish:

• active plasmid transfer at a specific time;
• functional resistance caused by one detected gene;
• persistence phenotype from transcriptomics alone;
• a complete epidemiological transmission route;
• clinical risk;
• patient treatment response.

Independent validation should match the claim being made. Useful options can include:

• culture and AST;
• targeted PCR;
• isolate WGS;
• long-read sequencing;
• plasmid reconstruction;
• conjugation experiments;
• time-series sampling;
• phenotypic survival assays;
• targeted expression validation;
• functional perturbation.

Validation is strongest when it directly tests the biological relationship proposed by the genomic analysis.

From Gene Detection to Mechanism-Aware AMR Surveillance

The next stage of AMR surveillance is unlikely to depend on replacing one technology with another. It is more likely to depend on linking several evidence layers.

From ARG Detection to ARG–Host Context

ARG abundance is valuable. But some research questions require the next layer: Which microbial genome carries the determinant, and in what genomic context? That is where strain-resolved assembly, long-read context, or single-cell genomic association can add value.

From Clone Tracking to Clone + Mobile-Element Tracking

A resistant clone can spread. A resistance plasmid can also spread independently of that clone. Mechanism-aware surveillance may therefore need to track:

• bacterial lineage;
• plasmid lineage;
• resistance cassette;
• host range;
• genomic context.

From Cultured Isolates to Broader Community Sampling

Cultured isolates remain essential. Culture-independent methods can broaden access to organisms that are difficult to recover or are not the dominant cultured taxa in a sample. The goal should not be “capture the entire community without bias.” No method can guarantee that. The goal is to reduce a known sampling blind spot.

From Population Average to Subpopulation Heterogeneity

Three different questions should remain separate: Genome: What genetic potential is present? Transcriptome: What molecular state is active under the measured condition? Phenotype: What does the organism actually do under antimicrobial exposure? A strong AMR study uses the layer that matches the conclusion. For broader microbiology applications within the same site, see Spatial Omics Solutions for Microbiology.

Cell-Resolved Microbial Genomics at CD Genomics

A useful project starts by identifying the unresolved link in the current data.

Start With the Surveillance Question

Examples include:

• Which host carries this ARG?
• Is strain resolution required?
• Which host carries this plasmid?
• Are uncultured or low-abundance populations central to the question?
• Is the problem genomic carriage or transcriptional state?
• Is a phenotype already available?

This prevents unnecessary assays.

Select the Minimum Necessary Data Layers

A genomic track may include:

• individual-cell genome recovery;
• SAG assembly and QC;
• taxonomic assignment;
• strain comparison;
• ARG annotation;
• mobile-element annotation;
• candidate host–element association.

A transcriptomic track may include:

• cell-resolved expression matrices;
• transcriptional clustering;
• differential expression;
• stress- or resistance-associated pathway analysis;
• state-composition comparison across experimental conditions.

These tracks can be used independently or combined when the biological question truly requires both genotype and state.

Typical Deliverables

Depending on project scope, deliverables may include:

• raw sequencing files;
• SAG assemblies where applicable;
• assembly and QC metrics;
• taxonomic annotations;
• ARG annotations;
• mobile-element annotations;
• strain-comparison outputs;
• cell-by-gene expression matrices;
• transcriptional clustering;
• differential-expression results;
• pathway or functional annotations;
• visualization-ready figures;
• documented analysis outputs.

Service-specific feasibility should be assessed for the sample and research question. Published study counts, genome-completeness values, or cell-recovery numbers should not be presented as universal expectations. Researchers can review our Microbial Single-Cell Sequencing Services for the genomic and transcriptomic tracks available for research projects.

If existing WGS or metagenomic data identify a resistance signal but cannot resolve who carries it, which mobile element contains it, or which microbial states respond under a defined condition, the next step is to identify the missing evidence layer. Share the sample type, microbial system, resistance question, and current data limitation when planning the project. The study can then be configured around the smallest set of assays needed to resolve that uncertainty.

Frequently Asked Questions

Conclusion: The Next Step Is Better Context, Not More Data by Default

NARMS 2026–2030 makes one direction clear: AMR surveillance is moving toward deeper genomic interpretation, better data integration, broader microbiome context, and new technologies that can improve how resistance is tracked. That does not make single-cell microbial sequencing the next mandatory surveillance platform. The more useful question is: What relationship remains unresolved after the methods already in use? If the missing information is phenotype, use phenotypic evidence. If it is isolate genome structure, use WGS. If it is community breadth, use metagenomics. If it is long genomic context, consider long reads. If it is cell-level host association or transcriptional heterogeneity, cell-resolved microbial omics may add a valuable layer.

The future of AMR surveillance will depend on combining these evidence types carefully—and validating each conclusion at the level it claims to explain.

References

1. U.S. Food and Drug Administration, Centers for Disease Control and Prevention, U.S. Department of Agriculture. The National Antimicrobial Resistance Monitoring System Strategic Plan for 2026–2030. 2026.
2. Centers for Disease Control and Prevention. Antimicrobial Resistance Threats in the United States, 2021–2022. Published July 2024.
3. Abramova A, Karkman A, Bengtsson-Palme J. Metagenomic assemblies tend to break around antibiotic resistance genes. BMC Genomics. 2024;25:959. DOI: 10.1186/s12864-024-10876-0.
4. Bloemen B, Gand M, Ringenier M, et al. Overcoming challenges in metagenomic AMR surveillance with nanopore sequencing: a case study on fluoroquinolone resistance. Frontiers in Microbiology. 2025;16:1614301. DOI: 10.3389/fmicb.2025.1614301.
5. Olsen NS, Riber L. Metagenomics as a Transformative Tool for Antibiotic Resistance Surveillance: Highlighting the Impact of Mobile Genetic Elements with a Focus on the Complex Role of Phages. Antibiotics. 2025;14(3):296. DOI: 10.3390/antibiotics14030296.
6. Zhang Y, Xue B, Mao Y, et al. High-throughput single-cell sequencing of activated sludge microbiome. Environmental Science and Ecotechnology. 2025;23:100493. DOI: 10.1016/j.ese.2024.100493.
7. Ye L, Wu Y, Guo J, et al. Elucidation of population-based bacterial adaptation to antimicrobial treatment by single-cell sequencing analysis of the gut microbiome of a hospital patient. mSystems. 2026;11(2):e01631-24. DOI: 10.1128/msystems.01631-24.
8. Wang H, Wu X, Xu J, et al. Proline mitigates antibiotic resistance evolution induced by ciprofloxacin at environmental concentrations. Journal of Hazardous Materials. 2025;489:137561. DOI: 10.1016/j.jhazmat.2025.137561.
9. Monte DFM, Thakur S. One Health transmission of plasmid-mediated antimicrobial resistance: genomic insights at the interface of food, animals, humans and the environment. JAC-Antimicrobial Resistance. 2026;8(4):dlag130. DOI: 10.1093/jacamr/dlag130.

Research Use and Trust Statement

This article discusses AMR surveillance and microbial omics for research use.

• Services are intended for research use only and are not intended for clinical diagnosis or individual treatment decisions.
• NARMS is an independent U.S. federal surveillance collaboration. Discussion of its strategic plan does not imply endorsement of CD Genomics or any specific single-cell technology.
• NARMS does not specifically recommend single-cell microbial sequencing in its 2026–2030 strategic plan.
• Single-cell microbial genomics should complement, not replace, AST, isolate WGS, long-read sequencing, or metagenomics when those methods already answer the research question.
• ARG–host and plasmid–host associations should be interpreted according to genome recovery, assembly quality, contamination, and the strength of independent evidence.
• Host–plasmid association does not by itself prove horizontal transfer.
• Transcriptomic state does not by itself establish antimicrobial resistance, tolerance, or persistence.
• Published cell counts, genome recovery rates, ARG counts, or other quantitative results are study-specific observations and should not be treated as universal technical specifications.
• Human-derived, animal-derived, food-associated, and environmental research materials should be collected and studied under the submitter's applicable ethical, biosafety, consent, transport, and institutional requirements.