Machine learning methods to identify markers and predict antimicrobial resistance in <i>Escherichia coli</i>
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Antimicrobial resistant strains of pathogenic Escherichia coli are a burden on the healthcare system, causing longer hospital stays and increased treatment costs compared to nonresistant strains. With whole genome sequencing almost ubiquitous in the analyses of outbreak and surveillance samples , in silico methods for feature identification can be faster and cheaper than traditional wet-lab methods. In this study, machine learning (ML) classification methods were used to predict antimicrobial resistance (AMR) and identify novel genomic markers of resistance. A total of 4300 E. coli whole genome sequences with laboratory-derived susceptible, intermediate, or resistant (SIR) data for 34 antimicrobials were collected. Three models—gradient boosted decision trees, support vector machines (SVMs), and artificial neural networks (ANNs)—were trained using genome subsequences ( k-mers) of length 11 to classify unknown isolates as SIR for each antimicrobial. The models achieved high average accuracies (93.6%, 92.7%, and 92.8%, respectively) for our dataset, outperforming database methods including AMRFinderPlus (63.9%) and ResFinder (75.7%). Tested on two smaller independent datasets, the models’ average accuracies were 81.6% (XGB), 79.9% (SVM), and 81.2% (ANN), while ResFinder’s average accuracy was 94.7%. An advantage of ML models over database methods is that they can identify novel markers of resistance, which is a key advantage for surveillance and research. As more genomic and AMR data become publicly available, these models are expected to further improve in performance and utility.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it