Assembly-Quality Contamination Screening and Its Impact on Antimicrobial Resistance Predictions for Typhoid Fever Treatment

This repository accompanies a reanalysis of the Global Typhoid Genomics Consortium (GTGC) dataset, testing whether assembly-quality screening measurably changes reported antimicrobial resistance (AMR) prevalence in Salmonella Typhi, and whether any such change reflects a genuine difference in genotype-call accuracy rather than an artefact of exclusion.

Structure

├── notebooks/
│   └── STyphi_CoLab_Final.ipynb   # Full analysis pipeline
├── data/
│   ├── gtgc_raw_input.csv         # GTGC line list (13,003 genomes, 57 countries)
│   └── styphi_amrnet_export.json  # AMRnet cross-check export
├── results/
│   ├── gtgc_vs_amr_results_summary.csv
│   ├── table1_arm_a_vs_b.csv          # Manuscript Table 1: Arm A vs Arm B prevalence
│   ├── table2_negative_control.csv    # Manuscript Table 2: placebo negative control
│   ├── table4_amrnet_concordance.csv  # Manuscript Table 3: AMRnet concordance
│   └── table5_ast_concordance.csv     # Manuscript Table 4: lab-confirmed AST validation
└── figures/
    ├── fig1_dataset_characterisation.* # Geographic and temporal distribution
    ├── fig2_arm_comparison.*           # Prevalence shift significance and negative control
    ├── fig3_amrnet_concordance.*       # Concordance with AMRnet by flag status
    └── fig4_ast_concordance.*          # Concordance with lab-confirmed phenotype

Each figure is provided as PDF, PNG, and TIFF.

Pipeline summary

A contamination-risk flag was constructed from genome length (outside the 2.5th–97.5th percentile), N50 (<20,000), contig count (>200), and the consortium's own exclude_assembly indicator. Resistance prevalence was compared between the full dataset (Arm A, n=13,003) and the flagged-excluded subset (Arm B, n=12,166) across six drug-class outcomes using chi-square tests, odds ratios, and Benjamini-Hochberg correction. A randomly permuted negative control (seed=42) was used to distinguish genuine signal from exclusion-of-any-subset artefacts. Calls were cross-checked against AMRnet and against lab-confirmed phenotypic AST data from the GTGC consortium's Mykrobe validation resource.

Full methodological detail is in the manuscript; the notebook in notebooks/ reproduces every reported statistic and figure.

Data provenance

• Primary dataset: fetched programmatically from the GTGC consortium's public repository. File integrity was recorded via SHA-256 hash at time of fetch (prefix: 9ec9addb3ed63e62).
• Independent cross-check: AMRnet, an independently curated global AMR surveillance resource.
• Phenotypic validation: the GTGC consortium's Mykrobe validation resource (lab-confirmed antimicrobial susceptibility testing data).
• No raw sequencing reads were generated, accessed, or deposited as part of this work — all inputs are pre-existing, publicly available consortium outputs.

Reproducing the analysis

The notebook was run in Google Colaboratory using Python 3.12.13 with pandas, NumPy, SciPy, statsmodels, matplotlib, and seaborn. Open notebooks/STyphi_CoLab_Final.ipynb in Colab or a local Jupyter environment with these packages installed; input data paths assume the repository's folder structure above.

Citation

A full citation with DOI will be added here upon publication.

License

Code in this repository is released under the MIT License. Data files are derived from publicly available consortium resources (GTGC, AMRnet, Mykrobe) — see Data provenance above for original sources and their respective usage terms.