AMR_Predictor

Machine learning methods to predict the anti-microbial resistance of Salmonella.

If you would like to use prebuilt models against your own genomes

1. Clone repository (run git clone https://github.com/superphy/AMR_Predictor.git)
2. Download anaconda or miniconda (python 3.7), instructions for that are here
3. Install dependecies: run conda env create -f data/envi.yaml

If you do not have your own MIC labels and would like to use ours (from NCBI SRA May 2019) skip to step 5

If you only want predictions with no evaluations, remove predict/mic_labels.xlsx and skip to step 6

If you have your own mic labels, proceed with step 4

4. Name your labels as mic_labels.xlsx and replace predict/mic_labels.xlsx

In mic_labels.xlsx the names of the genomes need to be in a column titled run and the MIC values need to be in columns labeled like MIC_AMP, MIC_AMC, etc

See predict/mic_labels.xlsx for acceptable MIC formats

5. Run snakemake -s predict/mic_clean.smk
6. Place genomes in predict/genomes/raw
7. Run snakemake -j X -s predict/predict.smk where X is the number of cores you wish to use
8. View results in predict/results.csv or predictions in predict/predictions.csv

If you would like to run all of the tests

1. Clone repository (run git clone https://github.com/superphy/AMR_Predictor.git)

2. Download anaconda or miniconda (python 3.7), instructions for that are here

3. Install dependecies: run conda env create -f data/envi.yaml

4. Move public genomes into AMR_Predictor/data/genomes/raw

5. Move grdi genomes in AMR_Predictor/data/grdi_genomes/raw (optional, but remove grdi rules from Snakefile)

6. Start conda environment: run source activate skmer

7. Run the following command, where 'X' is the number of cores you wish to use

   snakemake -j X

8. Run the following command to run all of the tests

   snakemake -s src/run_tests.smk or snakemake -s src/run_XGB_SVM_tests_slurm.smk && hyperas.smk(if using slurm)

9. Run all_data_figures.py all to save all the results as figures

Figures can be found in figures/

To find the results of an individual test, run result_grabber.py --help

10. If you would like to annotate the genomes and map back the top features:

run model.py with the -i flag, e.g. python src/model.py -x public -f 1000 -a AMP -i

Set the parameters in annotation/annotate.smk

run snakemake -j X -s annotation/annotate.smk

The resulting annotations and the location of the most import regions to the machine learning models can be found in annotation/

Manually Running Tests

If you do not want to run all tests in step 8 above, run src/model.py --help to see how to run the model for a specific set of parameters.