def bootstrap_profiles(profiles,
alpha=0.05,
repetitions=1000,
randfunc=randint):
"""Performs bootstrapping over the sample 'profiles'
Inputs:
profiles: list of profiles
alpha: defines the confidence interval as 1 - alpha
repetitions: number of bootstrap iterations
randfunc: random function for generate the bootstrap samples
Returns:
profile: the bootstrapped profile of the profiles list
sample_mean: the bootstrap mean of the amount shared
sample_stdev: the bootstrap standard deviation of the amount shared
ci: the confidence interval for the bootstrap mean
"""
length = len(profiles)
normalize_profiles(profiles)
boot_shared = []
boot_profiles = []
for i in range(repetitions):
# Construct the bootstrap sample
resample = [profiles[randfunc(0, length)] for j in range(length)]
profile = compare_profiles(resample)
# Store the amount shared
boot_shared.append(1.0 - profile['not_shared'])
# Store the result profile
boot_profiles.append(profile)
# Convert data to a numpy array
boot_shared = array(boot_shared)
# Get the mean and the standard deviation of the shared data
sample_mean = mean(boot_shared)
sample_stdev = std(boot_shared)
# Compute the confidence interval for the bootstrapped data
# using bootstrap percentile interval
ci = quantile(boot_shared, [alpha / 2, 1 - (alpha / 2)])
# Compute the bootstrapped profile of the profiles list
profile = compare_profiles(profiles)
return profile, 1.0 - profile['not_shared'], sample_stdev, ci
def bootstrap_profiles(profiles, alpha=0.05, repetitions=1000,
randfunc=randint):
"""Performs bootstrapping over the sample 'profiles'
Inputs:
profiles: list of profiles
alpha: defines the confidence interval as 1 - alpha
repetitions: number of bootstrap iterations
randfunc: random function for generate the bootstrap samples
Returns:
profile: the bootstrapped profile of the profiles list
sample_mean: the bootstrap mean of the amount shared
sample_stdev: the bootstrap standard deviation of the amount shared
ci: the confidence interval for the bootstrap mean
"""
length = len(profiles)
normalize_profiles(profiles)
boot_shared = []
boot_profiles = []
for i in range(repetitions):
# Construct the bootstrap sample
resample = [profiles[randfunc(0, length)] for j in range(length)]
profile = compare_profiles(resample)
# Store the amount shared
boot_shared.append(1.0 - profile['not_shared'])
# Store the result profile
boot_profiles.append(profile)
# Convert data to a numpy array
boot_shared = array(boot_shared)
# Get the mean and the standard deviation of the shared data
sample_mean = mean(boot_shared)
sample_stdev = std(boot_shared)
# Compute the confidence interval for the bootstrapped data
# using bootstrap percentile interval
ci = quantile(boot_shared, [alpha/2, 1-(alpha/2)])
# Compute the bootstrapped profile of the profiles list
profile = compare_profiles(profiles)
return profile, 1.0-profile['not_shared'], sample_stdev, ci
def add_alpha_diversity_values_to_mapping_file(metrics, alpha_sample_ids,
alpha_data, mapping_file_headers,
mapping_file_data, bins, method='equal',
missing_value_name='N/A'):
"""add 3 columns in the mapping file representing the alpha diversity data
Inputs:
metrics: list of alpha diversity metrics
alpha_sample_ids: list of sample identifiers in the alpha diversity data
alpha_data: alpha diversity data, as many columns as metrics and as many
rows as elments in alpha_sample_ids
mapping_file_headers: list of headers for the metadata mapping file
mapping_file_data: metadata mapping file data
bins: bins to classify the alpha diversity data
method: binning method selection, the options are 'equal' and 'quantile'.
'equal', will get you equally spaced limits and 'quantile' will assign the
limits using quantiles, using the selected number of bins.
missing_value_name: string to place for the sample ids in the mapping file
but not in the alpha diversity data
Output:
mapping_file_headers: modified headers mapping file headers, including three
new columns per metric i. e. for chao1 the new fields would be:
'chao1_alpha', 'chao1_alpha_norm' and 'chao1_alpha_bin'
mapping_file_data: extended data including the alpha diversity values,
normalized values and bins
"""
norm = lambda x, x_min, x_max: (x-x_min)/(x_max-x_min)
# data will be modified and returned so get your own copy
new_mapping_file_data = deepcopy(mapping_file_data)
new_mapping_file_headers = deepcopy(mapping_file_headers)
# regular levels assigned based on equally spaced bins
overall_probs = [i/bins for i in range(1, bins)]
# if we are using the average method the levels are equal to the probs list
if method == 'equal':
levels = overall_probs
for index, metric in enumerate(metrics):
# get the alpha diversity value for the metric being evaluated
data = [[row[index]] for row in alpha_data]
metric_max = max(data)[0]
metric_min = min(data)[0]
# add headers for each metric
new_mapping_file_headers.append('{0}_alpha'.format(metric))
new_mapping_file_headers.append('{0}_normalized_alpha'.format(metric))
new_mapping_file_headers.append('{0}_alpha_label'.format(metric))
# when using the quantile method the levels change depending on the
# metric being used; hence the calculation and normalization of the data
if method == 'quantile':
levels = quantile([norm(element[0], metric_min, metric_max)
for element in data], overall_probs)
# get the normalized value of diversity and the tag for each value
for value in data:
norm_value = norm(value[0], metric_min, metric_max)
value.append(norm_value)
value.append(_get_level(norm_value, levels, 'bin'))
# iterate using the mapping file instead of using the alpha diversity
# data because more often that you will like you will have more samples
# in the mapping file than in your downstream analysis data
for row in new_mapping_file_data:
try:
data_index = alpha_sample_ids.index(row[0])
# data fields should be strings
row.extend(map(str, data[data_index]))
except ValueError:
row.extend([missing_value_name, missing_value_name,
missing_value_name])
return new_mapping_file_data, new_mapping_file_headers
def add_alpha_diversity_values_to_mapping_file(metrics,
alpha_sample_ids,
alpha_data,
mapping_file_headers,
mapping_file_data,
bins,
method='equal',
missing_value_name='N/A'):
"""add 3 columns in the mapping file representing the alpha diversity data
Inputs:
metrics: list of alpha diversity metrics
alpha_sample_ids: list of sample identifiers in the alpha diversity data
alpha_data: alpha diversity data, as many columns as metrics and as many
rows as elments in alpha_sample_ids
mapping_file_headers: list of headers for the metadata mapping file
mapping_file_data: metadata mapping file data
bins: bins to classify the alpha diversity data
method: binning method selection, the options are 'equal' and 'quantile'.
'equal', will get you equally spaced limits and 'quantile' will assign the
limits using quantiles, using the selected number of bins.
missing_value_name: string to place for the sample ids in the mapping file
but not in the alpha diversity data
Output:
mapping_file_headers: modified headers mapping file headers, including three
new columns per metric i. e. for chao1 the new fields would be:
'chao1_alpha', 'chao1_alpha_norm' and 'chao1_alpha_bin'
mapping_file_data: extended data including the alpha diversity values,
normalized values and bins
"""
norm = lambda x, x_min, x_max: (x - x_min) / (x_max - x_min)
# data will be modified and returned so get your own copy
new_mapping_file_data = deepcopy(mapping_file_data)
new_mapping_file_headers = deepcopy(mapping_file_headers)
# regular levels assigned based on equally spaced bins
overall_probs = [i / bins for i in range(1, bins)]
# if we are using the average method the levels are equal to the probs list
if method == 'equal':
levels = overall_probs
for index, metric in enumerate(metrics):
# get the alpha diversity value for the metric being evaluated
data = [[row[index]] for row in alpha_data]
metric_max = max(data)[0]
metric_min = min(data)[0]
# add headers for each metric
new_mapping_file_headers.append('{0}_alpha'.format(metric))
new_mapping_file_headers.append('{0}_normalized_alpha'.format(metric))
new_mapping_file_headers.append('{0}_alpha_label'.format(metric))
# when using the quantile method the levels change depending on the
# metric being used; hence the calculation and normalization of the
# data
if method == 'quantile':
levels = quantile(
[norm(element[0], metric_min, metric_max) for element in data],
overall_probs)
# get the normalized value of diversity and the tag for each value
for value in data:
norm_value = norm(value[0], metric_min, metric_max)
value.append(norm_value)
value.append(_get_level(norm_value, levels, 'bin'))
# iterate using the mapping file instead of using the alpha diversity
# data because more often that you will like you will have more samples
# in the mapping file than in your downstream analysis data
for row in new_mapping_file_data:
try:
data_index = alpha_sample_ids.index(row[0])
# data fields should be strings
row.extend(map(str, data[data_index]))
except ValueError:
row.extend([
missing_value_name, missing_value_name, missing_value_name
])
return new_mapping_file_data, new_mapping_file_headers