Merge branch 'bugfix/conditional_prob_data' into 'master'
Conditional probability data update See merge request caimira/caimira!499
This commit is contained in:
Andre Henriques
commit
b42870dac2
8 changed files with 69 additions and 78 deletions
|
|
@ -193,7 +193,7 @@ class ConcentrationModel(BaseRequestHandler):
|
|||
)
|
||||
# Re-generate the report with the conditional probability of infection plot
|
||||
if self.get_cookie('conditional_plot'):
|
||||
form.conditional_probability_plot = True if self.get_cookie('conditional_plot') == '1' else False
|
||||
form.conditional_probability_viral_loads = True if self.get_cookie('conditional_plot') == '1' else False
|
||||
self.clear_cookie('conditional_plot') # Clears cookie after changing the form value.
|
||||
|
||||
report_task = executor.submit(
|
||||
|
|
|
|||
|
|
@ -15,7 +15,6 @@ DEFAULTS = {
|
|||
'precise_activity': '{}',
|
||||
'calculator_version': NO_DEFAULT,
|
||||
'ceiling_height': 0.,
|
||||
'conditional_probability_plot': False,
|
||||
'conditional_probability_viral_loads': False,
|
||||
'CO2_fitting_result': '{}',
|
||||
'exposed_coffee_break_option': 'coffee_break_0',
|
||||
|
|
|
|||
|
|
@ -31,7 +31,6 @@ class VirusFormData(FormData):
|
|||
arve_sensors_option: bool
|
||||
precise_activity: dict
|
||||
ceiling_height: float
|
||||
conditional_probability_plot: bool
|
||||
conditional_probability_viral_loads: bool
|
||||
CO2_fitting_result: dict
|
||||
floor_area: float
|
||||
|
|
@ -497,7 +496,6 @@ def baseline_raw_form_data() -> typing.Dict[str, typing.Union[str, float]]:
|
|||
'air_changes': '',
|
||||
'air_supply': '',
|
||||
'ceiling_height': '',
|
||||
'conditional_probability_plot': '0',
|
||||
'conditional_probability_viral_loads': '0',
|
||||
'exposed_coffee_break_option': 'coffee_break_4',
|
||||
'exposed_coffee_duration': '10',
|
||||
|
|
|
|||
|
|
@ -171,25 +171,14 @@ def calculate_report_data(form: VirusFormData, model: models.ExposureModel, exec
|
|||
expected_new_cases = np.array(model.expected_new_cases()).mean()
|
||||
exposed_presence_intervals = [list(interval) for interval in model.exposed.presence_interval().boundaries()]
|
||||
|
||||
if (model.data_registry.virological_data['virus_distributions'][form.virus_type]['viral_load_in_sputum'] == ViralLoads.COVID_OVERALL.value # type: ignore
|
||||
and form.conditional_probability_plot): # Only generate this data if covid_overall_vl_data is selected.
|
||||
|
||||
viral_load_in_sputum: models._VectorisedFloat = model.concentration_model.infected.virus.viral_load_in_sputum
|
||||
viral_loads, pi_means, lower_percentiles, upper_percentiles = manufacture_conditional_probability_data(model, prob)
|
||||
|
||||
uncertainties_plot_src = img2base64(_figure2bytes(uncertainties_plot(prob, viral_load_in_sputum, viral_loads,
|
||||
pi_means, lower_percentiles, upper_percentiles)))
|
||||
conditional_probability_data = {key: value for key, value in
|
||||
zip(('viral_loads', 'pi_means', 'lower_percentiles', 'upper_percentiles'),
|
||||
(viral_loads, pi_means, lower_percentiles, upper_percentiles))}
|
||||
vl_dist = list(np.log10(viral_load_in_sputum))
|
||||
|
||||
else:
|
||||
uncertainties_plot_src = None
|
||||
conditional_probability_data = None
|
||||
vl = model.concentration_model.virus.viral_load_in_sputum
|
||||
if isinstance(vl, np.ndarray): vl_dist = list(np.log10(model.concentration_model.virus.viral_load_in_sputum))
|
||||
else: vl_dist = np.log10(model.concentration_model.virus.viral_load_in_sputum)
|
||||
conditional_probability_data = None
|
||||
uncertainties_plot_src = None
|
||||
if (form.conditional_probability_viral_loads and
|
||||
model.data_registry.virological_data['virus_distributions'][form.virus_type]['viral_load_in_sputum'] == ViralLoads.COVID_OVERALL.value): # type: ignore
|
||||
# Generate all the required data for the conditional probability plot
|
||||
conditional_probability_data = manufacture_conditional_probability_data(model, prob)
|
||||
# Generate the matplotlib image based on the received data
|
||||
uncertainties_plot_src = img2base64(_figure2bytes(uncertainties_plot(prob, conditional_probability_data)))
|
||||
|
||||
return {
|
||||
"model_repr": repr(model),
|
||||
|
|
@ -208,10 +197,9 @@ def calculate_report_data(form: VirusFormData, model: models.ExposureModel, exec
|
|||
"prob_hist_bins": list(prob_dist_bins),
|
||||
"prob_probabilistic_exposure": prob_probabilistic_exposure,
|
||||
"expected_new_cases": expected_new_cases,
|
||||
"uncertainties_plot_src": uncertainties_plot_src,
|
||||
"CO2_concentrations": CO2_concentrations,
|
||||
"vl_dist": vl_dist,
|
||||
"conditional_probability_data": conditional_probability_data,
|
||||
"uncertainties_plot_src": uncertainties_plot_src,
|
||||
}
|
||||
|
||||
|
||||
|
|
@ -235,7 +223,6 @@ def generate_permalink(base_url, get_root_url, get_root_calculator_url, form: V
|
|||
|
||||
|
||||
def conditional_prob_inf_given_vl_dist(
|
||||
data_registry: DataRegistry,
|
||||
infection_probability: models._VectorisedFloat,
|
||||
viral_loads: np.ndarray,
|
||||
specific_vl: float,
|
||||
|
|
@ -247,7 +234,9 @@ def conditional_prob_inf_given_vl_dist(
|
|||
upper_percentiles = []
|
||||
|
||||
for vl_log in viral_loads:
|
||||
# Probability of infection corresponding to a certain viral load value in the distribution
|
||||
specific_prob = infection_probability[np.where((vl_log-step/2-specific_vl)*(vl_log+step/2-specific_vl)<0)[0]] #type: ignore
|
||||
|
||||
pi_means.append(specific_prob.mean())
|
||||
lower_percentiles.append(np.quantile(specific_prob, 0.05))
|
||||
upper_percentiles.append(np.quantile(specific_prob, 0.95))
|
||||
|
|
@ -259,69 +248,74 @@ def manufacture_conditional_probability_data(
|
|||
exposure_model: models.ExposureModel,
|
||||
infection_probability: models._VectorisedFloat
|
||||
):
|
||||
data_registry: DataRegistry = exposure_model.data_registry
|
||||
|
||||
min_vl = 2
|
||||
max_vl = 10
|
||||
step = (max_vl - min_vl)/100
|
||||
viral_loads = np.arange(min_vl, max_vl, step)
|
||||
specific_vl = np.log10(exposure_model.concentration_model.virus.viral_load_in_sputum)
|
||||
pi_means, lower_percentiles, upper_percentiles = conditional_prob_inf_given_vl_dist(data_registry, infection_probability, viral_loads,
|
||||
pi_means, lower_percentiles, upper_percentiles = conditional_prob_inf_given_vl_dist(infection_probability, viral_loads,
|
||||
specific_vl, step)
|
||||
|
||||
return list(viral_loads), list(pi_means), list(lower_percentiles), list(upper_percentiles)
|
||||
log10_vl_in_sputum = np.log10(exposure_model.concentration_model.infected.virus.viral_load_in_sputum)
|
||||
|
||||
return {
|
||||
'viral_loads': list(viral_loads),
|
||||
'pi_means': list(pi_means),
|
||||
'lower_percentiles': list(lower_percentiles),
|
||||
'upper_percentiles': list(upper_percentiles),
|
||||
'log10_vl_in_sputum': list(log10_vl_in_sputum),
|
||||
}
|
||||
|
||||
|
||||
def uncertainties_plot(infection_probability: models._VectorisedFloat,
|
||||
viral_load_in_sputum: models._VectorisedFloat,
|
||||
viral_loads: models._VectorisedFloat,
|
||||
pi_means: models._VectorisedFloat,
|
||||
lower_percentiles: models._VectorisedFloat,
|
||||
upper_percentiles: models._VectorisedFloat):
|
||||
conditional_probability_data: dict):
|
||||
|
||||
fig, axes = plt.subplots(2, 3,
|
||||
viral_loads: list = conditional_probability_data['viral_loads']
|
||||
pi_means: list = conditional_probability_data['pi_means']
|
||||
lower_percentiles: list = conditional_probability_data['lower_percentiles']
|
||||
upper_percentiles: list = conditional_probability_data['upper_percentiles']
|
||||
log10_vl_in_sputum: list = conditional_probability_data['log10_vl_in_sputum']
|
||||
|
||||
fig, ((axs00, axs01, axs02), (axs10, axs11, axs12)) = plt.subplots(nrows=2, ncols=3, # type: ignore
|
||||
gridspec_kw={'width_ratios': [5, 0.5] + [1],
|
||||
'height_ratios': [3, 1], 'wspace': 0},
|
||||
sharey='row',
|
||||
sharex='col')
|
||||
|
||||
# Type hint for axs
|
||||
axs: np.ndarray = np.array(axes)
|
||||
axs01.axis('off')
|
||||
axs11.axis('off')
|
||||
axs12.axis('off')
|
||||
|
||||
for y, x in [(0, 1)] + [(1, i + 1) for i in range(2)]:
|
||||
axs[y, x].axis('off')
|
||||
axs01.set_visible(False)
|
||||
|
||||
axs[0, 1].set_visible(False)
|
||||
axs00.plot(viral_loads, np.array(pi_means), label='Predictive total probability')
|
||||
axs00.fill_between(viral_loads, np.array(lower_percentiles), np.array(upper_percentiles), alpha=0.1, label='5ᵗʰ and 95ᵗʰ percentile')
|
||||
|
||||
axs[0, 0].plot(viral_loads, np.array(pi_means)/100, label='Predictive total probability')
|
||||
axs[0, 0].fill_between(viral_loads, np.array(lower_percentiles)/100, np.array(upper_percentiles)/100, alpha=0.1, label='5ᵗʰ and 95ᵗʰ percentile')
|
||||
axs02.hist(infection_probability, bins=30, orientation='horizontal')
|
||||
axs02.set_xticks([])
|
||||
axs02.set_xticklabels([])
|
||||
axs02.set_facecolor("lightgrey")
|
||||
|
||||
axs[0, 2].hist(infection_probability, bins=30, orientation='horizontal')
|
||||
axs[0, 2].set_xticks([])
|
||||
axs[0, 2].set_xticklabels([])
|
||||
axs[0, 2].set_facecolor("lightgrey")
|
||||
highest_bar = axs02.get_xlim()[1]
|
||||
axs02.set_xlim(0, highest_bar)
|
||||
|
||||
highest_bar = axs[0, 2].get_xlim()[1]
|
||||
axs[0, 2].set_xlim(0, highest_bar)
|
||||
|
||||
axs[0, 2].text(highest_bar * 0.5, 0.5,
|
||||
rf"$\bf{np.round(np.mean(infection_probability), 1)}$%", ha='center', va='center')
|
||||
axs[1, 0].hist(np.log10(viral_load_in_sputum),
|
||||
axs02.text(highest_bar * 0.5, 50,
|
||||
"$P(I)=$\n" + rf"$\bf{np.round(np.mean(infection_probability), 1)}$%", ha='center', va='center')
|
||||
axs10.hist(log10_vl_in_sputum,
|
||||
bins=150, range=(2, 10), color='grey')
|
||||
axs[1, 0].set_facecolor("lightgrey")
|
||||
axs[1, 0].set_yticks([])
|
||||
axs[1, 0].set_yticklabels([])
|
||||
axs[1, 0].set_xticks([i for i in range(2, 13, 2)])
|
||||
axs[1, 0].set_xticklabels(['$10^{' + str(i) + '}$' for i in range(2, 13, 2)])
|
||||
axs[1, 0].set_xlim(2, 10)
|
||||
axs[1, 0].set_xlabel('Viral load\n(RNA copies)', fontsize=12)
|
||||
axs[0, 0].set_ylabel('Conditional Probability\nof Infection', fontsize=12)
|
||||
axs10.set_facecolor("lightgrey")
|
||||
axs10.set_yticks([])
|
||||
axs10.set_yticklabels([])
|
||||
axs10.set_xticks([i for i in range(2, 13, 2)])
|
||||
axs10.set_xticklabels(['$10^{' + str(i) + '}$' for i in range(2, 13, 2)])
|
||||
axs10.set_xlim(2, 10)
|
||||
axs10.set_xlabel('Viral load\n(RNA copies)', fontsize=12)
|
||||
axs00.set_ylabel('Conditional Probability\nof Infection', fontsize=12)
|
||||
|
||||
axs[0, 0].text(9.5, -0.01, '$(i)$')
|
||||
axs[1, 0].text(9.5, axs[1, 0].get_ylim()[1] * 0.8, '$(ii)$')
|
||||
axs[0, 2].set_title('$(iii)$', fontsize=10)
|
||||
axs00.text(9.5, -0.01, '$(i)$')
|
||||
axs10.text(9.5, axs10.get_ylim()[1] * 0.8, '$(ii)$')
|
||||
axs02.set_title('$(iii)$', fontsize=10)
|
||||
|
||||
axs[0, 0].legend()
|
||||
axs00.legend()
|
||||
return fig
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -1,6 +1,6 @@
|
|||
function on_report_load(conditional_probability_plot) {
|
||||
function on_report_load(conditional_probability_viral_loads) {
|
||||
// Check/uncheck uncertainties image generation
|
||||
document.getElementById('conditional_probability_plot').checked = conditional_probability_plot
|
||||
document.getElementById('conditional_probability_viral_loads').checked = conditional_probability_viral_loads
|
||||
}
|
||||
|
||||
/* Generate the concentration plot using d3 library. */
|
||||
|
|
@ -1164,14 +1164,14 @@ function display_rename_column(bool, id) {
|
|||
else document.getElementById(id).style.display = 'none';
|
||||
}
|
||||
|
||||
function conditional_probability_plot(value, is_generated) {
|
||||
function conditional_probability_viral_loads(value, is_generated) {
|
||||
// If the image was previously generated, there is no need to reload the page.
|
||||
if (value && is_generated == 1) {
|
||||
document.getElementById('conditional_probability_div').style.display = 'block'
|
||||
}
|
||||
else if (value && is_generated == 0) {
|
||||
document.getElementById('label_conditional_probability_plot').innerHTML = `<span id="loading_spinner" class="spinner-border spinner-border-sm mr-2 mt-0" role="status" aria-hidden="true"></span>Loading...`;
|
||||
document.getElementById('conditional_probability_plot').setAttribute('disabled', true);
|
||||
document.getElementById('label_conditional_probability_viral_loads').innerHTML = `<span id="loading_spinner" class="spinner-border spinner-border-sm mr-2 mt-0" role="status" aria-hidden="true"></span>Loading...`;
|
||||
document.getElementById('conditional_probability_viral_loads').setAttribute('disabled', true);
|
||||
document.cookie = `conditional_plot= 1; path=/`;
|
||||
window.location.reload();
|
||||
}
|
||||
|
|
|
|||
|
|
@ -495,7 +495,7 @@
|
|||
</div>
|
||||
|
||||
<div class="form-check d-none">
|
||||
<input type="checkbox" id="conditional_probability_plot" class="tabbed form-check-input" name="conditional_probability_plot" value="0" disabled>
|
||||
<input type="checkbox" id="conditional_probability_viral_loads" class="tabbed form-check-input" name="conditional_probability_viral_loads" value="0" disabled>
|
||||
</div>
|
||||
|
||||
<span id="training_limit_error" class="red_text" hidden>Conference/Training activities limited to 1 infected<br></span>
|
||||
|
|
|
|||
|
|
@ -15,7 +15,7 @@
|
|||
|
||||
</head>
|
||||
|
||||
<body id="body" onload="on_report_load({{ form.conditional_probability_plot | int }})">
|
||||
<body id="body" onload="on_report_load({{ form.conditional_probability_viral_loads | int }})">
|
||||
|
||||
<!-- MODEL REPR - Available in the developer tools once the report is generated. Useful to re-create the model using an interpreter that has CAiMIRA installed:
|
||||
|
||||
|
|
@ -214,13 +214,13 @@
|
|||
draw_histogram("prob_inf_hist", {{ prob_inf }}, {{ prob_inf_sd }});
|
||||
</script>
|
||||
<br>
|
||||
{% if model.data_registry.virological_data['virus_distributions'][form.virus_type]['viral_load_in_sputum'] == 'Ref: Viral load - covid_overal_vl_data' %}
|
||||
{% if model.data_registry.virological_data['virus_distributions'][form.virus_type]['viral_load_in_sputum'] == 'Ref: Viral load - covid overal viral load data' %}
|
||||
<div class="form-check">
|
||||
<input type="checkbox" id="conditional_probability_plot" class="tabbed form-check-input" name="conditional_probability_plot" value="1" onClick="conditional_probability_plot(this.checked, {{ form.conditional_probability_plot | int }});">
|
||||
<label id="label_conditional_probability_plot" for="conditional_probability_plot" class="form-check-label col-sm-12">Generate full uncertainty data (as function of the viral load)</label>
|
||||
<input type="checkbox" id="conditional_probability_viral_loads" class="tabbed form-check-input" name="conditional_probability_viral_loads" value="1" onClick="conditional_probability_viral_loads(this.checked, {{ form.conditional_probability_viral_loads | int }});">
|
||||
<label id="label_conditional_probability_viral_loads" for="conditional_probability_viral_loads" class="form-check-label col-sm-12">Generate full uncertainty data (as function of the viral load)</label>
|
||||
</div>
|
||||
{% endif %}
|
||||
{% if form.conditional_probability_plot %}
|
||||
{% if form.conditional_probability_viral_loads %}
|
||||
<div id="conditional_probability_div">
|
||||
<img src= "{{ uncertainties_plot_src }}" />
|
||||
<div class="ml-5">
|
||||
|
|
|
|||
|
|
@ -72,7 +72,7 @@ def test_conditional_prob_inf_given_vl_dist(data_registry, baseline_exposure_mod
|
|||
specific_vl = np.log10(mc_model.concentration_model.infected.virus.viral_load_in_sputum)
|
||||
step = 8/100
|
||||
actual_pi_means, actual_lower_percentiles, actual_upper_percentiles = (
|
||||
report_generator.conditional_prob_inf_given_vl_dist(data_registry, infection_probability, viral_loads, specific_vl, step)
|
||||
report_generator.conditional_prob_inf_given_vl_dist(infection_probability, viral_loads, specific_vl, step)
|
||||
)
|
||||
|
||||
assert np.allclose(actual_pi_means, expected_pi_means, atol=0.002)
|
||||
|
|
|
|||
Loading…
Reference in a new issue