Adding functionality for negative phenotype annotation to Bauer et. Al's BOQA A single diagnosis (full run through of computing each marginal given a fixed configuration of annotations) currently takes ~1.2 seconds.
To recreate results: To generate patients: python generate_patient_pairs.py --out_path generated_patients_neg -N 100 --generate PATIENTS --imprecision --noise 0.5 --negative_phenotypes python generate_patient_pairs.py --out_path generated_patients_neg -N 100 --generate PATIENTS --imprecision --noise 0.5
To run respective methods on each set of patients: submit_(k|ic_sampling|sampling|p_sampling).sh directory_with_patients output_folder
To run nofreqs on each set of patients: python run_net.py --patient_path directory_with_patients --out_path output_folder
hpo.py, omim.py are adapted from github.com/buske/patient-similarity
generate_patient_pairs.py is adapted from github.com/talf301/matchingsim