Model specific configurations

Serialization

EMOD

EMOD allows serialization and to use an alternative base calibration.

Serialization allows the user to run a simulation, save their status and results at a specified point in time, and start another simulation (i.e. with modified campaigns) that continues from the point saved.
In EMOD, this is often applied to first run the burnin the long initial simulation run, during which population immunity is established, followed by the pickup which defines the often typical scenario explorations.

The following parameters need editing when running launch_sim.py:

  • To run a burnin simulation :

    • Set emod_step = ‘burnin’ .
    • The experiment ID for this burnin simulation will be needed when running the pickup simulation (and is saved automatically in folder, and also printed to console).

  • To run a pickup simulation(s):

    • In latest version of the framework, submitting a burnin simulation will automatically create and run a
      corresponding pickup simulation via launch_pickup.sh
    • In the files, emod_step is set to pickup and emod_serialized_id automatically to <burnin experiment id>
    • Note, pickup simulations cannot run without an attached experiment id (for EMOD).
    • The launch_pickup.sh includes the following lines after the required SBATCH specifications:

    ``` ml singularity/3.8.1
    ml python/3.8.4 source /projects/b1139/environments/mmmpy_venv/bin/activate

     cd /projects/b1139/emod-om/model-characterization
 python launch_sim.py -d <job_directory>/<exp_name>_burnin80 --suite <SUITE_Name> --expname <exp_name> --emodstep pickup --serializedid <emod_serialized_id> --models EMOD MalSim OpenMalaria

    ```

  • To run a customized (new) pickup simulation(s):

    • Use the generated launch_pickup.sh as a template:
    • Modify the expname to specify a new folder to store the simulation files and change emodstep to new_pickup
    • Edit the launch_sim.py as required, or alternatively use launch_simulations.py with the corresponding csv

    • Use launch_sim.py python launch_sim.py -d <job_directory>/<exp_name>_burnin80 --suite <SUITE_Name> --expname <exp_name_new> --emodstep new_pickup --serializedid <emod_serialized_id> --models EMOD MalSim OpenMalaria

    • In the current version of the framework the new simulation experiments has to remain within the same directory (Suite) as the simulation experiment ‘picked up from’.

  • To run a burnin and pickup simulation (without serializing, as selected as default for testing):

    • Set emod_step = None, or if running from terminal, do not specify emod_step, which will default to None

Using alternative base calibration

EMOD

This section is under construction…

OpenMalaria

Model variants
OpenMalaria allows to take into account model and parameter uncertainty through varying model variants). However, for simplicity in MultiMalModPy, only the base model is included.

Parameter sampling

malariasimulation

malariasimulation allows paramter sampling.

In malariasimulation, accounting for parameter uncertainty to better reflect potential variability in the simulation results has been integrated by drawing parameter values from posterior distributions around defined parameter sets.

Via the set_parameter_draw function up to 1000 different parameter sets for 32 simulation parameters can be selected. These 32 parameters correspond to: fixed state transitions, immunity boost grace periods, immunity decay rates, immunity reducing probability of detection, infectivity towards mosquitoes, mortality, probability of clinical infection, probability of pre-erythrocytic infection, and probability of severe infection.

In MultiMalModPy, seed values are paired with the parameter_draw value.

For a single seed, varying parameter sets from in total 1000 sets are drawn and a new draw is made each time new simulations run. A breakdown of each parameter that is affected in the parameter draw, with parameter type, description, default value, maximum and minimum is shown in the core parameter table for malariasimulation.

Importation

EMOD

Importation in EMOD is setup using the add_outbreak_individual function from emodpy-malaria.
The specific python code snippet, as used in MultiMalModPy here, is shown below: 

ncases_per_year_per1000pop = (exp.emod_importation_rate * exp.emod_pop_size) / 1000
add_outbreak_individual(camp, start_day=1, target_num_individuals=ncases_per_year_per1000pop,
                        repetitions=-1, timesteps_between_repetitions=365)

OpenMalaria

Importation in OpenMalaria is setup using importedInfections: The specific XML chunk, as used in MultiMalModPy here, is shown below: 

<importedInfections>
    <timed>
        <rate time="0" value="@importation_rate@"/>
    </timed>
</importedInfections>

Entomology mode

EMOD

Relevant code chunks are under EMOD_functions.py here and under EMOD/utils.py here.

‘Forced’ EIR uses the add_scheduled_input_eir function from emodpy-malaria.

## If forced EIR, treated as campaign in EMOD
if row.entomology_mode == 'forced':
    set_inputEIR_seasonality(camp, float(row.model_input_em), row.seasonality)

if entomology_mode == 'dynamic':
    task.common_assets.add_directory(os.path.join(manifest.input_dir, "example_weather"), relative_path="climate")

They also require different sweeping, defined in config_sweep_builder (and config_sweep_builder_pickup) see code. You can find details on parameter sweeping for building scenarios under simulation guide/scenarios

OpenMalaria

entomology_mode in OpenMalaria defines the feedback from human infectiousness to subsequent Entomological Infection Rate (EIR).
entomology_mode in OpenMalaria is setup by literally defining the ‘mode of entomology’: The specific XML chunk, as used in MultiMalModPy here, is shown below: 

<entomology mode="@entomology_mode@" name="Namawala" scaledAnnualEIR="@eir@">