Reproducible bioinformatics portfolio
AMR Insight
Genotype-aware exploration of recorded ciprofloxacin resistance in Escherichia coli from public isolate records.
Loading the research snapshot…
Rates describe this selected database snapshot only; they are not population prevalence.
Yearly points carry 95% Wilson confidence intervals and the isolate count behind them. Years holding fewer than 30 isolates are drawn hollow and listed below; their percentages move on a handful of records and should not be read as a trend.
Explore the included records
Countries
Specimen categories
Type to search, then choose as many countries or specimen categories as you need. Selected values appear as chips you can remove individually.
| BioSample | Assembly | Country | Year | Specimen | Phenotype | AMR genotypes |
|---|
Explore a genotype profile
This displays association with recorded phenotype labels. It is not a patient risk score and does not recommend an antibiotic.
Model association score
—
Modelled association between this isolate profile and a recorded resistant
label in the training snapshot. It is not a patient risk, not a probability of
treatment failure, and not a clinical threshold.
Held-out performance
Coefficients use the transformed feature scale. Their sign shows model association, not causation or isolated biological importance.
Study design and boundaries
Research question. Can isolate metadata and selected AMRFinderPlus genotype indicators distinguish recorded ciprofloxacin-resistant from susceptible E. coli isolates?
Inclusion. Exact E. coli records with a unique BioSample accession, valid collection year, and one unambiguous ciprofloxacin phenotype (R or S). Intermediate, conflicting, and missing phenotypes are excluded.
Model. Class-weighted logistic regression with broad source metadata, collection year, AMR genotype count, quinolone-resistance markers, selected QRDR substitutions, and beta-lactamase-family indicators. The newest viable years are held out.
Limits. Submitter-supplied AST methods and breakpoints may differ across laboratories and time. Submission patterns strongly shape the dataset. Genotype calls may correlate with study origin, laboratory practice, lineage, geography, and time. Strong discrimination is not causal or clinical validity.
Intended use. Portfolio research, reproducible bioinformatics, and hypothesis generation—not diagnosis, treatment selection, or population surveillance.