https://amr.poppunk.net/ is a tool to predict antimicrobial resistance in S.pneumoniae. It applies linear models on genetic sequence data to return a probability of resistance to the five antibiotics
- Penicillin
- Chloramphenicol
- Erythromycin
- Tetracycline
- Trimethoprim/Sulfamethoxazole
Files in FASTA format can be uploaded via a drag and drop area. Submission of multiple files at a time is possible. Files should cover the whole genome to ensure the models can be applied correctly.
Results are displayed in a table, giving the probability of resistance to the respective antibiotic.
The table can be downloaded in CSV format.
Don't use this tool for clinical practice.
The models applied are logistic ElasticNet models trained on data from the USA and South Africa from the GPS database.
Input features that are generated from the FASTA files are unitigs, which are nucleotide sequences of variable length.
The models have been tested on independent datasets from Massachusetts and Maela (Thailand). The predicted binary resistance status (at a probability threshold of 0.5) was tested against the true recorded phenotype and Balanced Accuracies were calculated:
| Antibiotic | Massachusetts | Maela |
|---|---|---|
| Penicillin | 0.933 | 0.836 |
| Chlorampehnicol | 0.819 | |
| Erythromycin | 0.959 | 0.961 |
| Tetracycline | 0.953 | 0.940 |
| Trimethoprim/Sulfamethoxazole | 0.954 | 0.837 |
The website was build using React.js. The backend code can be found here.
This tool was developed by Marie Gronemeyer and John Lees.