February 18, 2025

Detecting AMR from Low Bacterial Load Clinical Samples

Antimicrobial resistance (AMR) is projected to cause more deaths than cancer by 2050, yet the standard diagnostic method — microbiological culture — takes days and frequently returns negative results in patients who clearly have infections. Detecting AMR genes from low bacterial load samples like blood, cerebrospinal fluid, and bronchoalveolar lavage demands a fundamentally different approach. This webinar summary covers a rapid mNGS workflow that combines host depletion, isothermal DNA amplification, and Oxford Nanopore sequencing to identify both pathogens and antimicrobial resistance genes from clinical samples in under three hours.

Watch the full webinar:

Key Takeaways

A three-step workflow — host depletion, rolling circle amplification, and nanopore sequencing — identified pathogens from clinical samples in under 3 hours with only 30 minutes of hands-on time
Charge-based filtration depleted host cells in 5 minutes without chemicals or enzymes, enabling processing of low-biomass samples
Rolling circle amplification (RCA) from whole bacterial cells outperformed multiple displacement amplification for generating sufficient DNA for sequencing
The workflow identified Mycoplasma pneumoniae in a culture-negative bronchopneumonia case and Haemophilus influenzae in a case dismissed as contamination
AMR gene detection requires at least 15x genome coverage and phenotype prediction tools — genotype alone does not reliably predict antibiotic susceptibility

The Clinical Problem — AMR Detection When Culture Fails

Consider a patient admitted with rapidly deteriorating sepsis. Empiric antibiotics are administered immediately — the standard of care when every hour of delay increases mortality. Blood cultures are drawn, but results take two to five days. In many cases, the cultures return negative despite clear clinical evidence of infection. The empiric antibiotics may or may not cover the actual pathogen, and without identification of the organism and its resistance profile, clinicians are treating blind.

Existing molecular alternatives have limitations. Cell-free DNA (cfDNA) metagenomic assays provide broad pathogen detection but cost thousands of dollars per test and require 24 to 72 hours. Targeted PCR panels deliver results in one to two hours but test only approximately 20 pathogens — they are not hypothesis-free. The unmet need: a rapid, affordable test that identifies both the pathogen and its AMR genes from the same clinical sample within a single shift.

Building a 3-Hour Sample-to-Answer Workflow

The workflow presented in this webinar addresses three sequential challenges: removing the overwhelming host DNA background, amplifying the tiny amount of remaining microbial DNA, and sequencing it fast enough to be clinically actionable.

Step 1 — Host Depletion via Charge-Based Filtration

In a typical blood sample, host cells outnumber microbial cells by orders of magnitude. Without depletion, sequencing produces almost entirely human reads — useless data for pathogen identification. The workflow uses the Devin Filter from Micronbrane Medical, which employs zwitterionic charge-based filtration to selectively bind mammalian nucleated cells while allowing bacteria, fungi, and viruses to pass through. The process takes five minutes with no chemicals or enzymes. Compared to differential lysis (60-80 minutes, digests cell-free microbial DNA) and hybridization capture (1.5-3 hours, requires probe design), charge-based filtration was the fastest option compatible with point-of-care use.

Step 2 — Microbial DNA Enrichment Without PCR

After host depletion, the remaining microbial DNA is typically in the picogram to low nanogram range — far below the input requirements for most sequencing library preparations. The workflow uses isothermal whole genome amplification rather than PCR to avoid amplification bias.

Two methods were compared: multiple displacement amplification (MDA), commonly used for linear DNA in forensic applications, and rolling circle amplification (RCA), designed for circular genomes like plasmids. Since bacterial genomes are also circular, the hypothesis was that RCA would be more efficient.

RCA directly from whole bacterial cells (skipping DNA isolation entirely) yielded the highest DNA output — also the simplest workflow. At clinically relevant concentrations of Staphylococcus aureus (10 CFU), the amplified DNA yielded approximately 10x genome coverage.

Step 3 — Rapid Library Prep and Nanopore Sequencing

The amplified DNA contains branched structures from isothermal amplification that must be fragmented before sequencing. The Oxford Nanopore Rapid Sequencing Kit uses transposon complexes to simultaneously fragment DNA and attach sequencing adapters in 15 minutes — compared to hours for ligation-based library preparation.

Sequencing was performed on the Flongle flow cell (126 pores), the smallest and most affordable nanopore device. Real-time basecalling means pathogen identification can begin within minutes of loading the flow cell. The total workflow — from clinical sample to first sequencing results — takes under three hours with approximately 30 minutes of hands-on time, using only a heat block and a microcentrifuge.

Clinical Validation — Pathogen ID in Culture-Negative Cases

Three clinical samples demonstrated the workflow’s diagnostic value.

Sputum from bronchopneumonia (22-year-old female). Microbiological culture was negative. Nanopore mNGS identified Mycoplasma pneumoniae — an organism that is notoriously difficult to culture and frequently missed by standard diagnostics.

Urine from urinary tract infection (79-year-old female). Culture was positive for Pseudomonas aeruginosa as the most prevalent organism. The sequencing workflow correctly identified the same pathogen, confirming concordance with culture in positive cases.

Bronchoalveolar lavage from severe pneumonia (patient with hepatic complications). Culture grew organisms classified as contaminants. The sequencing workflow identified Haemophilus influenzae as the most prevalent pathogen — very likely the true cause of pneumonia that culture dismissed.

Antimicrobial Resistance Gene Detection from Sequencing Data

Identifying the pathogen is only half the clinical question. Determining which antibiotics will work requires detecting AMR genes. The workflow queried two databases: CARD (Comprehensive Antibiotic Resistance Database), which identifies both confirmed resistance genes and resistance-related genes, and NCBI AMRFinderPlus, which focuses more narrowly on genes with established resistance phenotypes.

For the Pseudomonas aeruginosa UTI sample (approximately 5-6x genome coverage), the CARD database returned nearly 20 hits while AMRFinderPlus returned 4. When compared to antibiotic sensitivity testing from culture, only cephalosporin resistance was correctly predicted by sequencing. Aminoglycoside resistance was predicted by the genotype but contradicted by phenotypic testing, which showed sensitivity.

For a Veillonella species detected in a respiratory sample (approximately 20x coverage), two resistance genes were identified — erythromycin and clindamycin resistance — though whether Veillonella was the true etiological agent remained uncertain.

Bridging the Gap Between Genotype and Phenotype

The AMR results reveal a critical limitation: genotype does not always equal phenotype. The presence of a resistance gene in the genome does not guarantee that the gene is expressed, functional, or clinically relevant. Conversely, resistance mechanisms not encoded by known genes (such as efflux pump upregulation or porin mutations) may be missed entirely by database queries.

Reliable AMR prediction from sequencing data requires at least 15x genome coverage (or greater than 80% de novo assembled genome), validated resistance gene databases, and machine learning-based phenotype prediction tools that go beyond simple gene presence/absence calls. Translating these results into clinician-friendly reports remains a prerequisite for replacing culture-based antibiotic sensitivity testing.

The PaRTI-Seq mNGS workflow from Micronbrane Medical integrates host depletion, optimized library preparation for low-biomass samples, and a cloud-based bioinformatic pipeline to streamline both pathogen identification and downstream resistance analysis.

About the Speaker

Tomas Chansky is a PhD student at the Institute of Molecular Biomedicine, Faculty of Medicine, Comenius University Bratislava, Slovakia. His research focuses on nucleic acid analysis for investigating the etiology and pathophysiology of infectious diseases, with emphasis on direct-from-blood metagenomics and nanopore sequencing for diagnosing sepsis-causing pathogens. He holds a master’s degree in clinical biochemistry from Comenius University.