Genomic Surveillance and Resistome in ICUs: Unraveling Klebsiella pneumoniae and Acinetobacter baumannii with an Interactive Dashboard

1. Introduction: The Challenge of Antimicrobial Resistance in Brazilian ICUs

Antimicrobial resistance (AMR) in opportunistic bacterial pathogens is one of the most severe threats to global public health. In the Intensive Care Unit (ICU), where immunocompromised patients with invasive devices face serious hospital-acquired infections, multidrug-resistant Gram-negative bacteria such as carbapenemase-producing (KPC) Klebsiella pneumoniae and carbapenem-resistant Acinetobacter baumannii (CRAB) lead to challenging clinical outcomes and a shortage of therapeutic options.

However, monitoring the spread of these pathogens with traditional phenotypic tests alone is no longer enough. We need to investigate the genomic architecture of circulating strains: which clones are expanding? Which resistance determinants sit on plasmids or integrons? How do chromosomal mutations affect last-line antimicrobials such as colistin?

In this article, we share the behind-the-scenes work and genomic findings of a large prospective multicenter study conducted in six tertiary hospitals in São Paulo, coordinated by Prof. Dr. Eduardo Alexandrino Servolo de Medeiros (Infectious Diseases Division / LEMC / UNIFESP), in which I worked directly on genomic curation, bioinformatic auditing and the development of analytical tools. To make this large dataset accessible, we built a complete analytical dashboard, available interactively right here on the blog!

2. The Multicenter Cohort and the Bioinformatics Pipeline

The study followed ICUs at six major hospitals in São Paulo State: Hospital São Paulo (HSP), Hospital Estadual de Diadema (HED), Hospital Municipal de Barueri (HMB), Hospital Geral de Pirajussara (HGPI), Hospital Geral de Guarulhos (HGG) and Hospital Geral de Pedreira (HGPE).

Starting from 290 initial raw samples from active surveillance (rectal swabs) and diagnostic cultures, we applied strict quality control (QC) and epidemiological inclusion criteria, consolidating a definitive cohort of 236 isolates with Whole Genome Sequencing (WGS) on the Illumina NextSeq 1000 platform:

The Decisive Role of Bioinformatic Curation: Practical Lessons

In routine clinical bioinformatics, automated "out-of-the-box" pipelines often produce pitfalls that only expert curation can uncover:

1. The mgrB trap in colistin resistance: in K. pneumoniae, mgrB inactivation is the main cause of polymyxin resistance. However, standard point-SNP alignment tools reported that most resistant isolates had no mgrB mutation. By inspecting base coverage, we found that 28 resistant isolates had 0% coverage of the mgrB gene — that is, they had undergone complete structural deletion or disruption by insertion sequences (IS). Once the artifact was corrected, 77.8% (35/45) of colistin-resistant isolates turned out to carry functional mgrB inactivation!
2. Detection and exclusion of chimeras/contamination: isolate HED-117-ACB stood out with 304 virulence genes (versus an average of 111 for Acinetobacter). An in-depth audit revealed complete biosynthesis pathways for pyocyanin, pyoverdine and type III secretion systems exclusive to Pseudomonas aeruginosa. It was pre-sequencing microbiological contamination that, had it not been audited, would have seriously distorted the group's publications.
3. International breakpoint harmonization (BrCAST/EUCAST vs. CLSI): we aligned reports between the Brazilian ANVISA standard (BrCAST v14.0) and the international standard (CLSI M100). For colistin in A. baumannii, for example, the "Susceptible" category does not exist in CLSI (MIC ≤ 2 µg/mL is classified as Intermediate), requiring dual reporting for compatibility with high-impact international journals.

3. Key Biological Findings Revealed by WGS

5. How to Run and Publish the Shiny Application

If you want to run the R Shiny version of the dashboard on your own machine or deploy it to shinyapps.io, the code is fully prepared and modularized.

Running locally in R / RStudio

# Install the required packages
install.packages(c("shiny", "bslib", "DT", "plotly", "rsconnect", "bsicons"))

# Run the application
shiny::runApp("app.R")

Publishing to shinyapps.io

Just insert your shinyapps.io credentials (Account → Tokens → Show):

library(rsconnect)

rsconnect::setAccountInfo(
  name   = "logikbioinfo",
  token  = "YOUR_TOKEN",
  secret = "YOUR_SECRET"
)

rsconnect::deployApp(
  appDir        = getwd(),
  appPrimaryDoc = "app.R",
  appName       = "amr-genomics-multicenter-unifesp"
)

6. Conclusion and LogikBioinfo Services

Integrating classical microbiology, next-generation sequencing and curated bioinformatics is today's gold standard for high-impact research and health surveillance. Cases such as elucidating the mgrB mechanism and identifying the NDM-1 cluster in Diadema show that detailed bioinformatic analysis goes far beyond generating raw tables: it turns nucleotide sequences into actionable epidemiological knowledge.