class MetagenomeCentricGeneClassifier:
def __init__(self, args, run=run, progress=progress):
self.run = run
self.progress = progress
A = lambda x: args.__dict__[x] if x in args.__dict__ else None
self.gene_coverages_data_file_path = A('data_file')
self.gene_detections_data_file_path = A('gene_detection_data_file')
self.profile_db_path = A('profile_db')
self.output_file_prefix = A('output_file_prefix')
self.alpha = A('alpha')
self.additional_layers_to_append = A('additional_layers_to_append')
self.samples_information_to_append = A('samples_information_to_append')
self.collection_name = A('collection_name')
self.bin_id = A('bin_id')
self.bin_ids_file_path = A('bin_ids_file')
self.exclude_samples = A('exclude_samples')
self.include_samples = A('include_samples')
self.profile_db = {}
self.coverage_values_per_nt = {}
self.gene_coverages = pd.DataFrame.empty
self.gene_detections = pd.DataFrame.empty
self.samples = {}
self.sample_detection_information_was_initiated = False
self.positive_samples = []
self.number_of_positive_samples = None
self.negative_samples = pd.DataFrame.empty
self.number_of_negative_samples = None
self.gene_class_information = pd.DataFrame.empty
self.samples_detection_information = pd.DataFrame.empty
self.gene_presence_absence_in_samples = pd.DataFrame.empty
self.gene_coverages_filtered = pd.DataFrame.empty
self.additional_description = ''
self.total_length = None
self.samples_coverage_stats_dicts_was_initiated = False
self.samples_coverage_stats_dicts = pd.DataFrame.empty
self.non_outlier_indices = {}
if self.exclude_samples:
# check that there is a file like this
filesnpaths.is_file_exists(self.exclude_samples)
self.samples_to_exclude = set([
l.split('\t')[0].strip()
for l in open(self.exclude_samples, 'rU').readlines()
])
if not self.samples_to_exclude:
raise ConfigError(
"You asked to exclude samples, but provided an empty list."
)
run.info(
'Excluding Samples',
'The following samples will be excluded: %s' %
self.samples_to_exclude,
)
else:
self.samples_to_exclude = set([])
if self.include_samples:
# check that there is a file like this
filesnpaths.is_file_exists(self.include_samples)
self.samples_to_include = set([
l.split('\t')[0].strip()
for l in open(self.include_samples, 'rU').readlines()
])
if not self.samples_to_include:
raise ConfigError(
"You provided an empty list of samples to include.")
run.info(
'Including Samples',
'The following samples will be included: %s' %
self.samples_to_include,
)
else:
self.samples_to_include = set([])
# run sanity check on all input arguments
self.sanity_check()
if self.profile_db_path is None:
# TODO: this will probably be removed because we don't save the coverage information in nucleotide level.
pass
else:
# load sample list and gene_coverage_dict from the merged profile db
args.init_gene_coverages = True
if self.collection_name:
self.summary = summarizer.ProfileSummarizer(args)
self.summary.init()
self.init_samples(self.summary.p_meta['samples'])
else:
self.profile_db = ProfileSuperclass(args)
self.init_samples(self.profile_db.p_meta['samples'])
self.profile_db.init_split_coverage_values_per_nt_dict()
self.profile_db.init_gene_level_coverage_stats_dicts()
self.coverage_values_per_nt = get_coverage_values_per_nucleotide(
self.profile_db.split_coverage_values_per_nt_dict,
self.samples)
# comply with the new design and get gene_coverages and gene_detection dicsts from
# gene_level_coverage_stats_dict.
gene_coverages, gene_detection = self.get_gene_coverages_and_gene_detection_dicts(
)
self.init_coverage_and_detection_dataframes(
gene_coverages, gene_detection)
# getting the total length of all contigs
self.total_length = self.profile_db.p_meta['total_length']
def get_gene_coverages_and_gene_detection_dicts(self):
gene_coverages = {}
gene_detection = {}
A = lambda x: self.profile_db.gene_level_coverage_stats_dict[
gene_callers_id][sample_name][x]
gene_caller_ids = list(
self.profile_db.gene_level_coverage_stats_dict.keys())
# populate gene coverage and detection dictionaries
if self.profile_db.gene_level_coverage_stats_dict:
for gene_callers_id in gene_caller_ids:
gene_coverages[gene_callers_id], gene_detection[
gene_callers_id] = {}, {}
for sample_name in self.profile_db.p_meta['samples']:
gene_coverages[gene_callers_id][sample_name] = A(
'mean_coverage')
gene_detection[gene_callers_id][sample_name] = A(
'detection')
return gene_coverages, gene_detection
def check_if_valid_portion_value(self, arg_name, arg_value):
""" Helper function to verify that an argument has a valid value for a non-zero portion (i.e. greater than zero and a max of 1)"""
if arg_value <= 0 or arg_value > 1:
raise ConfigError(
"%s value must be greater than zero and a max of 1, the value you supplied %s"
% (arg_name, arg_value))
def sanity_check(self):
"""Basic sanity check for class inputs"""
if self.profile_db_path is None and self.gene_coverages_data_file_path is None:
raise ConfigError(
"You must provide either a profile.db or a gene coverage self.gene_coverages_filtered data file"
)
if self.profile_db_path and self.gene_coverages_data_file_path:
raise ConfigError(
"You provided both a profile database and a gene coverage self.gene_coverages_filtered data file, you \
must provide only one or the other (hint: if you have a profile database, the use that"
)
# checking output file
filesnpaths.is_output_file_writable(self.output_file_prefix +
'-additional-layers.txt',
ok_if_exists=False)
# checking alpha
if not isinstance(self.alpha, float):
raise ConfigError("alpha value must be a type float.")
# alpha must be a min of 0 and smaller than 0.5
if self.alpha < 0 or self.alpha >= 0.5:
raise ConfigError(
"alpha must be a minimum of 0 and smaller than 0.5")
if self.collection_name:
if not self.profile_db_path:
raise ConfigError(
"You specified a collection name %s, but you provided a gene coverage self.gene_coverages_filtered data file \
collections are only available when working with a profile database."
% self.collection_name)
if self.exclude_samples and self.include_samples:
raise ConfigError(
"You cannot use both --include-samples and --exclude-samples! Please choose one."
)
def init_samples(self, samples_list):
""" Create the set of samples according to user input and store it in self.samples"""
samples = set(samples_list) - self.samples_to_exclude
if self.include_samples:
samples_to_include_that_are_not_there = self.samples_to_include - samples
if samples_to_include_that_are_not_there:
raise ConfigError(
"You requested to include some samples that are not in the profile database. Here are the samples in the profile database: %s. \
And here are the samples you requested, and that are not there: %s"
% (samples, samples_to_include_that_are_not_there))
samples = self.samples_to_include
self.samples = samples
def init_coverage_and_detection_dataframes(self, gene_coverages_dict,
gene_detection_dict):
""" Populate the following: self.gene_coverages, self.Ng, self.gene_detections.
Notice that this function could get as input either an object of ProfileSuperclass or of summarizer.Bin
"""
self.gene_coverages = pd.DataFrame.from_dict(gene_coverages_dict,
orient='index',
dtype=float)
self.Ng = len(self.gene_coverages.index)
self.gene_detections = pd.DataFrame.from_dict(gene_detection_dict,
orient='index',
dtype=float)
if self.include_samples or self.exclude_samples:
# Only include samples that the user want
self.gene_coverages = self.gene_coverages[list(self.samples)]
self.gene_detections = self.gene_detections[list(self.samples)]
def init_sample_detection_information(self):
""" Determine positive, negative, and ambiguous samples with the genome detection information
(--alpha, --genome-detection-uncertainty)
"""
# FIXME: some of the following variables are never used.
MCG_samples_information_table_name = 'MCG_classifier_samples_information'
MCG_samples_information_table_structure = [
'samples', 'presence', 'detection',
'number_of_taxon_specific_core_detected'
]
MCG_samples_information_table_types = ['str', 'bool', 'int', 'int']
# create an empty dataframe
samples_information = pd.DataFrame(
index=self.samples,
columns=MCG_samples_information_table_structure[1:])
positive_samples = []
negative_samples = []
self.progress.new("Setting presence/absence in samples")
num_samples, counter = len(self.samples), 1
detection = {}
for sample in self.samples:
if num_samples > 100 and counter % 100 == 0:
self.progress.update('%d of %d samples...' %
(counter, num_samples))
print("total length for %s is %s" % (sample, self.total_length))
print(
"the length of the vector: %s" %
len(self.coverage_values_per_nt[sample])
) # FIXME: after testing this module, delete this line. it is only here to make sure that anvio is not lying to us.
print(
"number of nucleotide positions with non zero coverage in %s is %s "
% (sample, np.count_nonzero(
self.coverage_values_per_nt[sample])))
detection[sample] = np.count_nonzero(
self.coverage_values_per_nt[sample]) / self.total_length
if detection[sample] >= 0.5 + self.alpha:
positive_samples.append(sample)
samples_information['presence'][sample] = True
elif detection[sample] <= 0.5 - self.alpha:
negative_samples.append(sample)
samples_information['presence'][sample] = False
else:
samples_information['presence'][sample] = None
samples_information['detection'][sample] = detection[sample]
counter += 1
self.progress.end()
self.positive_samples = positive_samples
self.number_of_positive_samples = len(self.positive_samples)
self.negative_samples = negative_samples
self.samples_detection_information = samples_information
self.run.warning('The number of positive samples is %s' %
self.number_of_positive_samples)
self.run.warning('The number of negative samples is %s' %
len(self.negative_samples))
self.sample_detection_information_was_initiated = True
def init_samples_coverage_stats_dict(self):
""" populate the samples_coverage_stats_dict."""
if not self.sample_detection_information_was_initiated:
self.init_sample_detection_information
self.samples_coverage_stats_dicts = pd.DataFrame(
index=self.samples,
columns=columns_for_samples_coverage_stats_dict)
num_samples, counter = len(self.samples), 1
self.progress.new(
"Finding nucleotide positions in samples with outlier coverage values"
)
for sample in self.positive_samples:
if num_samples > 100 and counter % 100 == 0:
self.progress.update('%d of %d samples...' %
(counter, num_samples))
# loop through positive samples
# get the non-zero non-outlier information
self.non_outlier_indices[
sample], self.samples_coverage_stats_dicts.loc[
sample, ] = get_non_outliers_information(
self.coverage_values_per_nt[sample])
self.run.info_single(
'The mean and std of non-outliers in sample %s are: %s, %s respectively'
%
(sample,
self.samples_coverage_stats_dicts['non_outlier_mean_coverage']
[sample],
self.samples_coverage_stats_dicts['non_outlier_coverage_std']
[sample]))
number_of_non_outliers = len(self.non_outlier_indices[sample])
self.run.info_single(
'The number of non-outliers is %s of %s (%.2f%%)' %
(number_of_non_outliers, self.total_length,
100.0 * number_of_non_outliers / self.total_length))
self.progress.end()
def plot_TS(self):
""" Creates a pdf file with the following plots for each sample the sorted nucleotide coverages \
(with a the outliers in red and non-outliers in blue), and a histogram of coverages for the non-outliers"""
# Creating a dircetory for the plots. If running on bins, each bin would be in a separate sub-directory
if not self.samples_coverage_stats_dicts_was_initiated:
self.init_samples_coverage_stats_dict()
additional_description = ''
if self.additional_description:
additional_description = '-' + self.additional_description
plot_dir = self.output_file_prefix + '-TS-plots' + '/'
os.makedirs(plot_dir, exist_ok=True)
self.progress.new(
'Saving figures of taxon specific distributions to pdf')
number_of_fininshed = 0
for sample in self.positive_samples:
coverages_pdf_output = plot_dir + sample + additional_description + '-coverages.pdf'
pdf_output_file = PdfPages(coverages_pdf_output)
v = self.coverage_values_per_nt[sample]
# Using argsort so we can use the non_oulier indices
sorting_indices = np.argsort(v)
# we would need the reverse of the sorting of the indices to create the x axis for the non-outliers
reverse_sorted_indices = np.zeros(len(sorting_indices))
reverse_sorted_indices[sorting_indices] = range(
len(reverse_sorted_indices))
# plotting the ordered coverage values (per nucleotide)
# the non-outliers are plotted in blue
# the outlier values are plotted in red
fig = plt.figure()
ax = fig.add_subplot(111, rasterized=True)
ax.set_xlabel = 'Nucleotide Number (ordered)'
ax.set_ylabel = r'$Nucleotide Coverage^2$'
x1 = range(
len(v)
) # FIXME: this shouldn't be in the loop (only here because I need to fix the mock data)
x2 = reverse_sorted_indices[self.non_outlier_indices[sample]]
y2 = v[self.non_outlier_indices[sample]]
# plot all in red
ax.semilogy(x1, v[sorting_indices], 'r.', rasterized=True)
# plot on top the non-outliers in blue
ax.semilogy(x2,
v[self.non_outlier_indices[sample]],
'b.',
rasterized=True)
fig.suptitle("%s - sorted coverage values with outliers" % sample)
plt.savefig(pdf_output_file, format='pdf')
plt.close()
# plotting a histogram of the non-outliers
# This would allow to see if they resemble a normal distribution
hist_range = (min(v[self.non_outlier_indices[sample]]),
max(v[self.non_outlier_indices[sample]]))
# computing the number of bins so that the width of a bin is ~1/4 of the standard deviation
# FIXME: need to make it so the bins are only of integers (so the smallest bin is of width 1
# and that bins are integers)
number_of_hist_bins = np.ceil(
(hist_range[1] - hist_range[0]) /
(self.samples_coverage_stats_dicts['non_outlier_coverage_std']
[sample] / 4)
).astype(
int
) # setting the histogram bins to be of the width of a quarter of std
fig = plt.figure()
ax = fig.add_subplot(111, rasterized=True)
ax.set_xlabel = 'Coverage'
ax.hist(v[self.non_outlier_indices[sample]],
number_of_hist_bins,
hist_range,
rasterized=True)
fig.suptitle("%s - histogram of non-outliers" % sample)
# adding the mean and std of the non-outliers as text to the plot
text_for_hist = u'$\mu = %d$\n $\sigma = %d$' %\
(self.samples_coverage_stats_dicts['non_outlier_mean_coverage'][sample],\
self.samples_coverage_stats_dicts['non_outlier_coverage_std'][sample])
ax.text(0.8,
0.9,
text_for_hist,
ha='center',
va='center',
transform=ax.transAxes)
plt.savefig(pdf_output_file, format='pdf')
plt.close()
# close the pdf file
pdf_output_file.close()
number_of_fininshed += 1
self.progress.update(
"Finished %d of %d" %
(number_of_fininshed, self.number_of_positive_samples))
self.progress.end()
def get_gene_classes(self):
""" The main process of this class - computes the class information for each gene"""
# need to start a new gene_class_information dict
# this is due to the fact that if the algorithm is ran on a list of bins then this necessary
self.gene_class_information = pd.DataFrame(
index=self.gene_coverages.index, columns=['gene_class'])
# set the presence/absence values for samples
self.init_sample_detection_information()
# find the taxon-specific genes for each sample
self.plot_TS()
def get_coverage_and_detection_dict(self, bin_id):
_bin = summarizer.Bin(self.summary, bin_id)
self.coverage_values_per_nt = get_coverage_values_per_nucleotide(
_bin.split_coverage_values_per_nt_dict, self.samples)
# getting the total length of all contigs
self.total_length = _bin.total_length
self.init_coverage_and_detection_dataframes(_bin.gene_coverages,
_bin.gene_detection)
def classify(self):
if self.collection_name:
bin_names_in_collection = self.summary.bin_ids
if self.bin_ids_file_path:
filesnpaths.is_file_exists(self.bin_ids_file_path)
bin_names_of_interest = [
line.strip()
for line in open(self.bin_ids_file_path).readlines()
]
missing_bins = [
b for b in bin_names_of_interest
if b not in bin_names_in_collection
]
if len(missing_bins):
raise ConfigError(
"Some bin names you declared do not appear to be in the collection %s. \
These are the bins that are missing: %s, these are the bins that are \
actually in your collection: %s" %
(self.collection_name, missing_bins,
bin_names_in_collection))
elif self.bin_id:
if self.bin_id not in bin_names_in_collection:
raise ConfigError("The bin you declared, %s, does not appear to be in the collection %s." \
% (self.bin_id, self.collection_name))
bin_names_of_interest = [self.bin_id]
else:
bin_names_of_interest = bin_names_in_collection
for bin_id in bin_names_of_interest:
self.run.info_single('Classifying genes in bin: %s' % bin_id)
self.get_coverage_and_detection_dict(bin_id)
self.additional_description = bin_id
self.get_gene_classes()
#self.save_gene_class_information_in_additional_layers(bin_id)
#self.save_samples_information(bin_id)
else:
# No collection provided so running on the entire detection table
self.get_gene_classes()
class MetagenomeCentricGeneClassifier:
def __init__(self, args, run=run, progress=progress):
self.run = run
self.progress = progress
A = lambda x: args.__dict__[x] if x in args.__dict__ else None
self.gene_coverages_data_file_path = A('data_file')
self.gene_detections_data_file_path = A('gene_detection_data_file')
self.profile_db_path = A('profile_db')
self.output_file_prefix = A('output_file_prefix')
self.alpha = A('alpha')
self.additional_layers_to_append = A('additional_layers_to_append')
self.samples_information_to_append = A('samples_information_to_append')
self.collection_name = A('collection_name')
self.bin_id = A('bin_id')
self.bin_ids_file_path = A('bin_ids_file')
self.exclude_samples = A('exclude_samples')
self.include_samples = A('include_samples')
self.profile_db = {}
self.coverage_values_per_nt = {}
self.gene_coverages = pd.DataFrame.empty
self.gene_detections = pd.DataFrame.empty
self.samples = {}
self.sample_detection_information_was_initiated = False
self.positive_samples = []
self.number_of_positive_samples = None
self.negative_samples = pd.DataFrame.empty
self.number_of_negative_samples = None
self.gene_class_df = pd.DataFrame.empty
self.samples_detection_information = pd.DataFrame.empty
self.gene_presence_absence_in_samples_initiated = False
self.gene_presence_absence_in_samples = pd.DataFrame.empty
self.gene_coverages_filtered = pd.DataFrame.empty
self.additional_description = ''
self.total_length = None
self.samples_coverage_stats_dicts_was_initiated = False
self.samples_coverage_stats_dicts = pd.DataFrame.empty
self.non_outlier_indices = {}
self.gene_coverage_stats_dict_of_dfs_initiated = False
self.gene_coverage_stats_dict_of_dfs = {}
self.gene_coverage_consistency_dict = {}
self.gene_coverage_consistency_dict_initiated = False
if self.exclude_samples:
# check that there is a file like this
filesnpaths.is_file_exists(self.exclude_samples)
self.samples_to_exclude = set([
l.split('\t')[0].strip()
for l in open(self.exclude_samples, 'rU').readlines()
])
if not self.samples_to_exclude:
raise ConfigError(
"You asked to exclude samples, but provided an empty list."
)
run.info(
'Excluding Samples',
'The following samples will be excluded: %s' %
self.samples_to_exclude,
)
else:
self.samples_to_exclude = set([])
if self.include_samples:
# check that there is a file like this
filesnpaths.is_file_exists(self.include_samples)
self.samples_to_include = set([
l.split('\t')[0].strip()
for l in open(self.include_samples, 'rU').readlines()
])
if not self.samples_to_include:
raise ConfigError(
"You provided an empty list of samples to include.")
run.info(
'Including Samples',
'The following samples will be included: %s' %
self.samples_to_include,
)
else:
self.samples_to_include = set([])
# run sanity check on all input arguments
self.sanity_check()
if self.profile_db_path is None:
# TODO: this will probably be removed because we don't save the coverage information in nucleotide level.
pass
else:
# load sample list and gene_coverage_dict from the merged profile db
args.init_gene_coverages = True
if self.collection_name:
self.summary = summarizer.ProfileSummarizer(args)
self.summary.init()
self.init_samples(self.summary.p_meta['samples'])
else:
self.profile_db = ProfileSuperclass(args)
self.init_samples(self.profile_db.p_meta['samples'])
self.profile_db.init_split_coverage_values_per_nt_dict()
self.profile_db.init_gene_level_coverage_stats_dicts(
outliers_threshold=2.5, populate_nt_level_coverage=True)
self.coverage_values_per_nt = get_coverage_values_per_nucleotide(
self.profile_db.split_coverage_values_per_nt_dict,
self.samples)
# comply with the new design and get gene_coverages and gene_detection dicsts from
# gene_level_coverage_stats_dict.
gene_coverages, gene_detection = self.get_gene_coverages_and_gene_detection_dicts(
)
self.init_coverage_and_detection_dataframes(
gene_coverages, gene_detection)
# getting the total length of all contigs
self.total_length = self.profile_db.p_meta['total_length']
def get_gene_coverages_and_gene_detection_dicts(self):
gene_coverages = {}
gene_detection = {}
A = lambda x: self.profile_db.gene_level_coverage_stats_dict[
gene_callers_id][sample_name][x]
gene_caller_ids = list(
self.profile_db.gene_level_coverage_stats_dict.keys())
# populate gene coverage and detection dictionaries
if self.profile_db.gene_level_coverage_stats_dict:
for gene_callers_id in gene_caller_ids:
gene_coverages[gene_callers_id], gene_detection[
gene_callers_id] = {}, {}
for sample_name in self.profile_db.p_meta['samples']:
gene_coverages[gene_callers_id][sample_name] = A(
'mean_coverage')
gene_detection[gene_callers_id][sample_name] = A(
'detection')
return gene_coverages, gene_detection
def check_if_valid_portion_value(self, arg_name, arg_value):
""" Helper function to verify that an argument has a valid value for a non-zero portion (i.e. greater than zero and a max of 1)"""
if arg_value <= 0 or arg_value > 1:
raise ConfigError(
"%s value must be greater than zero and a max of 1, the value you supplied %s"
% (arg_name, arg_value))
def sanity_check(self):
"""Basic sanity check for class inputs"""
if self.profile_db_path is None and self.gene_coverages_data_file_path is None:
raise ConfigError(
"You must provide either a profile.db or a gene coverage self.gene_coverages_filtered data file"
)
if self.profile_db_path and self.gene_coverages_data_file_path:
raise ConfigError(
"You provided both a profile database and a gene coverage self.gene_coverages_filtered data file, you \
must provide only one or the other (hint: if you have a profile database, the use that"
)
# checking output file
filesnpaths.is_output_file_writable(self.output_file_prefix +
'-additional-layers.txt',
ok_if_exists=False)
# checking alpha
if not isinstance(self.alpha, float):
raise ConfigError("alpha value must be a type float.")
# alpha must be a min of 0 and smaller than 0.5
if self.alpha < 0 or self.alpha >= 0.5:
raise ConfigError(
"alpha must be a minimum of 0 and smaller than 0.5")
if self.collection_name:
if not self.profile_db_path:
raise ConfigError(
"You specified a collection name %s, but you provided a gene coverage self.gene_coverages_filtered data file \
collections are only available when working with a profile database."
% self.collection_name)
if self.exclude_samples and self.include_samples:
raise ConfigError(
"You cannot use both --include-samples and --exclude-samples! Please choose one."
)
def init_samples(self, samples_list):
""" Create the set of samples according to user input and store it in self.samples"""
samples = set(samples_list) - self.samples_to_exclude
if self.include_samples:
samples_to_include_that_are_not_there = self.samples_to_include - samples
if samples_to_include_that_are_not_there:
raise ConfigError(
"You requested to include some samples that are not in the profile database. Here are the samples in the profile database: %s. \
And here are the samples you requested, and that are not there: %s"
% (samples, samples_to_include_that_are_not_there))
samples = self.samples_to_include
self.samples = samples
def init_coverage_and_detection_dataframes(self, gene_coverages_dict,
gene_detection_dict):
""" Populate the following: self.gene_coverages, self.Ng, self.gene_detections.
Notice that this function could get as input either an object of ProfileSuperclass or of summarizer.Bin
"""
self.gene_coverages = pd.DataFrame.from_dict(gene_coverages_dict,
orient='index',
dtype=float)
self.Ng = len(self.gene_coverages.index)
self.gene_detections = pd.DataFrame.from_dict(gene_detection_dict,
orient='index',
dtype=float)
if self.include_samples or self.exclude_samples:
# Only include samples that the user want
self.gene_coverages = self.gene_coverages[list(self.samples)]
self.gene_detections = self.gene_detections[list(self.samples)]
def init_sample_detection_information(self):
""" Determine positive, negative, and ambiguous samples with the genome detection information
(--alpha, --genome-detection-uncertainty)
"""
# FIXME: some of the following variables are never used.
MCG_samples_information_table_name = 'MCG_classifier_samples_information'
MCG_samples_information_table_structure = [
'samples', 'presence', 'detection',
'number_of_taxon_specific_core_detected'
]
MCG_samples_information_table_types = ['str', 'bool', 'int', 'int']
# create an empty dataframe
samples_information = pd.DataFrame(
index=self.samples,
columns=MCG_samples_information_table_structure[1:])
positive_samples = []
negative_samples = []
self.progress.new("Setting presence/absence in samples")
progress.update('...')
num_samples, counter = len(self.samples), 1
detection = {}
for sample in self.samples:
if num_samples > 100 and counter % 100 == 0:
self.progress.update('%d of %d samples...' %
(counter, num_samples))
print("total length for %s is %s" % (sample, self.total_length))
print(
"the length of the vector: %s" %
len(self.coverage_values_per_nt[sample])
) # FIXME: after testing this module, delete this line. it is only here to make sure that anvio is not lying to us.
print(
"number of nucleotide positions with non zero coverage in %s is %s "
% (sample, np.count_nonzero(
self.coverage_values_per_nt[sample])))
detection[sample] = np.count_nonzero(
self.coverage_values_per_nt[sample]) / self.total_length
samples_information['presence'][
sample] = get_presence_absence_information(
detection[sample], self.alpha)
if samples_information['presence'][sample]:
positive_samples.append(sample)
elif samples_information['presence'][sample] == False:
negative_samples.append(sample)
samples_information['detection'][sample] = detection[sample]
counter += 1
self.progress.end()
self.positive_samples = positive_samples
self.number_of_positive_samples = len(self.positive_samples)
self.negative_samples = negative_samples
self.samples_detection_information = samples_information
self.run.warning('The number of positive samples is %s' %
self.number_of_positive_samples)
self.run.warning('The number of negative samples is %s' %
len(self.negative_samples))
self.sample_detection_information_was_initiated = True
def init_samples_coverage_stats_dict(self):
""" populate the samples_coverage_stats_dict."""
if not self.sample_detection_information_was_initiated:
self.init_sample_detection_information()
self.samples_coverage_stats_dicts = pd.DataFrame(
index=self.samples,
columns=columns_for_samples_coverage_stats_dict)
num_samples, counter = len(self.samples), 1
self.progress.new(
"Finding nucleotide positions in samples with outlier coverage values"
)
progress.update('...')
for sample in self.positive_samples:
if num_samples > 100 and counter % 100 == 0:
self.progress.update('%d of %d samples...' %
(counter, num_samples))
# loop through positive samples
# get the non-outlier information
self.non_outlier_indices[
sample], self.samples_coverage_stats_dicts.loc[
sample, ] = get_non_outliers_information(
self.coverage_values_per_nt[sample])
self.run.info_single(
'The mean and std of non-outliers in sample %s are: %s, %s respectively'
%
(sample,
self.samples_coverage_stats_dicts['non_outlier_mean_coverage']
[sample],
self.samples_coverage_stats_dicts['non_outlier_coverage_std']
[sample]))
number_of_non_outliers = len(self.non_outlier_indices[sample])
self.run.info_single(
'The number of non-outliers is %s of %s (%.2f%%)' %
(number_of_non_outliers, self.total_length,
100.0 * number_of_non_outliers / self.total_length))
self.progress.end()
def plot_TS(self):
""" Creates a pdf file with the following plots for each sample the sorted nucleotide coverages \
(with a the outliers in red and non-outliers in blue), and a histogram of coverages for the non-outliers"""
# Creating a dircetory for the plots. If running on bins, each bin would be in a separate sub-directory
if not self.samples_coverage_stats_dicts_was_initiated:
self.init_samples_coverage_stats_dict()
additional_description = ''
if self.additional_description:
additional_description = '-' + self.additional_description
plot_dir = self.output_file_prefix + '-TS-plots' + '/'
os.makedirs(plot_dir, exist_ok=True)
self.progress.new(
'Saving figures of taxon specific distributions to pdf')
progress.update('...')
number_of_fininshed = 0
for sample in self.positive_samples:
coverages_pdf_output = plot_dir + sample + additional_description + '-coverages.pdf'
pdf_output_file = PdfPages(coverages_pdf_output)
v = self.coverage_values_per_nt[sample]
# Using argsort so we can use the non_oulier indices
sorting_indices = np.argsort(v)
# we would need the reverse of the sorting of the indices to create the x axis for the non-outliers
reverse_sorted_indices = np.zeros(len(sorting_indices))
reverse_sorted_indices[sorting_indices] = range(
len(reverse_sorted_indices))
# plotting the ordered coverage values (per nucleotide)
# the non-outliers are plotted in blue
# the outlier values are plotted in red
fig = plt.figure()
ax = fig.add_subplot(111, rasterized=True)
ax.set_xlabel = 'Nucleotide Number (ordered)'
ax.set_ylabel = r'$Nucleotide Coverage^2$'
x1 = range(
len(v)
) # FIXME: this shouldn't be in the loop (only here because I need to fix the mock data)
x2 = reverse_sorted_indices[self.non_outlier_indices[sample]]
y2 = v[self.non_outlier_indices[sample]]
# plot all in red
ax.semilogy(x1, v[sorting_indices], 'r.', rasterized=True)
# plot on top the non-outliers in blue
ax.semilogy(x2,
v[self.non_outlier_indices[sample]],
'b.',
rasterized=True)
fig.suptitle("%s - sorted coverage values with outliers" % sample)
plt.savefig(pdf_output_file, format='pdf')
plt.close()
# plotting a histogram of the non-outliers
# This would allow to see if they resemble a normal distribution
hist_range = (min(v[self.non_outlier_indices[sample]]),
max(v[self.non_outlier_indices[sample]]))
# computing the number of bins so that the width of a bin is ~1/4 of the standard deviation
# FIXME: need to make it so the bins are only of integers (so the smallest bin is of width 1
# and that bins are integers)
number_of_hist_bins = np.ceil(
(hist_range[1] - hist_range[0]) /
(self.samples_coverage_stats_dicts['non_outlier_coverage_std']
[sample] / 4)
).astype(
int
) # setting the histogram bins to be of the width of a quarter of std
fig = plt.figure()
ax = fig.add_subplot(111, rasterized=True)
ax.set_xlabel = 'Coverage'
ax.hist(v[self.non_outlier_indices[sample]],
number_of_hist_bins,
hist_range,
rasterized=True)
fig.suptitle("%s - histogram of non-outliers" % sample)
# adding the mean and std of the non-outliers as text to the plot
text_for_hist = u'$\mu = %d$\n $\sigma = %d$' %\
(self.samples_coverage_stats_dicts['non_outlier_mean_coverage'][sample],\
self.samples_coverage_stats_dicts['non_outlier_coverage_std'][sample])
ax.text(0.8,
0.9,
text_for_hist,
ha='center',
va='center',
transform=ax.transAxes)
plt.savefig(pdf_output_file, format='pdf')
plt.close()
# close the pdf file
pdf_output_file.close()
number_of_fininshed += 1
self.progress.update(
"Finished %d of %d" %
(number_of_fininshed, self.number_of_positive_samples))
self.progress.end()
def init_gene_presence_absence_in_samples(self):
gene_callers_id = list(
self.profile_db.gene_level_coverage_stats_dict.keys())
self.gene_presence_absence_in_samples = pd.DataFrame(
index=gene_callers_id, columns=self.samples)
num_samples, counter = len(self.samples), 1
self.progress.new('Computing gene presence/absence in samples')
progress.update('...')
for sample in self.samples:
if num_samples > 100 and counter % 100 == 0:
self.progress.update('%d of %d samples...' %
(counter, num_samples))
for gene_id in gene_callers_id:
self.gene_presence_absence_in_samples.loc[
gene_id, sample] = get_presence_absence_information(
self.profile_db.gene_level_coverage_stats_dict[gene_id]
[sample]['detection'], self.alpha)
self.gene_presence_absence_in_samples_initiated = True
self.progress.end()
def init_gene_coverage_stats_dict_of_dfs(self):
""" Convert gene_coverage_stats_dict to a dictionary of pandas dataframes
The reason to do this is that this way the gene parameters accross samples
could be used as numpy arrays. For example this allows to use the gene non-outlier
mean coverage accross samples as an array in order to perform regression
(see init_gene_coverage_consistency_information for example of usage).
"""
num_genes, counter = len(
self.profile_db.gene_level_coverage_stats_dict.keys()), 1
self.progress.new(
"Initializing gene coverage stats dictionary of dataframes")
progress.update('...')
for gene_id, coverage_stats in self.profile_db.gene_level_coverage_stats_dict.items(
):
if num_genes > 100 and counter % 100 == 0:
self.progress.update('%d of %d genes...' %
(counter, num_genes))
self.gene_coverage_stats_dict_of_dfs[
gene_id] = pd.DataFrame.from_dict(coverage_stats,
orient='index')
self.gene_coverage_stats_dict_of_dfs_initiated = True
self.progress.end()
def init_gene_coverage_consistency_information(self):
""" Perform orthogonal distance regression for each gene to determine coverage consistency.
The question that we are trying to ask is:
Do the non-outlier nt coverage of the gene in samlpes correlates to the non-outlier
nt coverage of the genome in samples?
The regression is performed only for positive samples.
For each gene, the regression is performed only according to samples in which
the gene is present (according to the detection critrea).
"""
if not self.samples_coverage_stats_dicts_was_initiated:
self.init_samples_coverage_stats_dict()
if not self.gene_presence_absence_in_samples_initiated:
self.init_gene_presence_absence_in_samples()
if not self.gene_coverage_stats_dict_of_dfs_initiated:
self.init_gene_coverage_stats_dict_of_dfs()
self.progress.new("Computing coverage consistency for all genes.")
progress.update('...')
num_genes, counter = len(
self.profile_db.gene_level_coverage_stats_dict.keys()), 1
for gene_id in self.profile_db.gene_level_coverage_stats_dict.keys():
if num_genes > 100 and counter % 100 == 0:
self.progress.update('%d of %d genes...' %
(counter, num_genes))
# samples in which the gene is present
_samples = self.gene_presence_absence_in_samples.loc[
gene_id, self.gene_presence_absence_in_samples.loc[
gene_id, ] == True].index
# mean and std of non-outlier nt in each sample
x = self.samples_coverage_stats_dicts.loc[
_samples, 'non_outlier_mean_coverage']
std_x = self.samples_coverage_stats_dicts.loc[
_samples, 'non_outlier_coverage_std']
if len(_samples) > 1:
# mean and std of non-outlier nt in the gene (in each sample)
y = self.gene_coverage_stats_dict_of_dfs[gene_id].loc[
_samples, 'non_outlier_mean_coverage']
std_y = self.gene_coverage_stats_dict_of_dfs[gene_id].loc[
_samples, 'non_outlier_coverage_std']
# performing the regression using ODR
_data = odr.RealData(list(x.values), list(y.values),
list(std_x.values), list(std_y.values))
_model = lambda B, c: B[0] * c
_odr = odr.ODR(_data, odr.Model(_model), beta0=[3])
odr_output = _odr.run()
# store results
self.gene_coverage_consistency_dict[gene_id] = {}
self.gene_coverage_consistency_dict[gene_id][
'slope'] = odr_output.beta[0]
self.gene_coverage_consistency_dict[gene_id][
'slope_std'] = odr_output.sd_beta[0]
self.gene_coverage_consistency_dict[gene_id][
'slope_precision'] = odr_output.sd_beta[
0] / odr_output.beta[0]
# compute R squered
f = lambda b: lambda _x: b * _x
R_squered = 1 - sum(
(np.apply_along_axis(f(odr_output.beta[0]), 0, x) -
y.values)**2) / sum((y - np.mean(y.values))**2)
# Check if converged
self.gene_coverage_consistency_dict[gene_id][
'R_squered'] = R_squered
if odr_output.stopreason[0] == 'Sum of squares convergence':
self.gene_coverage_consistency_dict[gene_id][
'converged'] = True
else:
self.gene_coverage_consistency_dict[gene_id][
'converged'] = False
self.gene_coverage_consistency_dict_initiated = True
self.progress.end()
def get_gene_specificity(self, gene_id):
""" return True for gene if it occurs in positive samples and doesn't occur in negative samples.
Ambiguous occurences are not counted as anything. This means that if a gene is ambiguously
occuring in a negative sample it could still be counted as "specific". It also means that
if a gene is only ambiguously occuring in positive samples then it would be considered
as "non-specific".
"""
if self.gene_class_df.loc[
gene_id,
'occurence_in_positive_samples'] > 1 and self.gene_class_df.loc[
gene_id, 'occurence_in_negative_samples'] == 0:
return True
else:
return False
# TODO: if there are no occurences of the gene at all, then we should maybe return None instead of False
def get_gene_coverage_consistency(self, gene_id):
""" return true if the gene's coverage is consistent accross positive samples, False otherwise."""
# TODO: make sure coverage_consistency_dict has been initiated
if self.gene_class_df.loc[gene_id,
'occurence_in_positive_samples'] == 0:
# if the gene doesn't occur in positive samlpes then there is no classification
return None
elif self.gene_class_df.loc[gene_id,
'occurence_in_positive_samples'] == 1:
# if the gene occurs only in one positive sample then return True.
# XXX: we might prefer to return None, we should consider this in the future.
return True
elif self.gene_coverage_consistency_dict[gene_id]['converged']:
# FIXME: this is where we use an arbitrary threshold again :-(
# if the slope precision is smaller than the threshold then the regression
# fit is considered accurate enough and the gene coverage is considered consistent.
return self.gene_coverage_consistency_dict[gene_id][
'slope_precision'] > 0.5
else:
# The regression didn't converege so the coverage is probably not consistent.
return False
def determine_if_gene_is_core(self, gene_id, gene_specificity):
""" return True for core gene, False for accessory gene
If the gene is specific to positive samples, then core would be considered if it
occurs in all positive samples. Otherwise it would be considered core if it
occurs in all positive AND all negative samples.
Ambiguous occurences of a gene are not considered (i.e. they are the same as absence).
"""
if gene_specificity:
# return True if the the gene occurs in all positive samples.
return self.gene_class_df.loc[
gene_id,
'occurence_in_positive_samples'] == len(self.positive_samples)
else:
# return True if the gene occurs in all positive AND all negative samples
return self.gene_class_df.loc[
gene_id, 'occurence_in_positive_and_negative_samples'] == len(
self.positive_samples) + len(self.negative_samples)
def init_gene_class_df(self):
""" generate dictionary with the class information per gene.
This dictionary could be later use to produce an additional-layer
text file for vizualization.
"""
# TODO: make sure gene presence absence was calculated
if not self.gene_coverage_consistency_dict_initiated:
self.init_gene_coverage_consistency_information()
# XXX: only negative and positive samples are used here
# ambiguous samples are ignored as if they were never
# there. This is not ideal, but is easy to do.
self.gene_class_df = pd.DataFrame(
index=list(self.profile_db.gene_level_coverage_stats_dict.keys()))
for gene_id in self.profile_db.gene_level_coverage_stats_dict.keys():
# determine the number of occurences in positive samples
self.gene_class_df.loc[gene_id,
'occurence_in_positive_samples'] = np.sum(
self.gene_presence_absence_in_samples.
loc[gene_id, self.positive_samples])
# determine the number of occurences in negative samples
self.gene_class_df.loc[gene_id,
'occurence_in_negative_samples'] = np.sum(
self.gene_presence_absence_in_samples.
loc[gene_id, self.negative_samples])
# set the occurence_in_positive_and_negative_samples
self.gene_class_df.loc[
gene_id,
'occurence_in_positive_and_negative_samples'] = self.gene_class_df.loc[
gene_id,
'occurence_in_positive_samples'] + self.gene_class_df.loc[
gene_id, 'occurence_in_negative_samples']
gene_specificity = self.get_gene_specificity(gene_id)
gene_coverage_consistency = self.get_gene_coverage_consistency(
gene_id)
# determine core accessory
gene_is_core = self.determine_if_gene_is_core(
gene_id, gene_specificity)
self.gene_class_df.loc[gene_id, 'specificity'] = gene_specificity
self.gene_class_df.loc[
gene_id, 'coverage_consistency'] = gene_coverage_consistency
self.gene_class_df.loc[gene_id, 'core'] = gene_is_core
self.gene_class_df.loc[gene_id, 'MCG_class'] = get_class_string(
gene_specificity, gene_coverage_consistency, gene_is_core)
def get_gene_classes(self):
""" The main process of this class - computes the class information for each gene"""
# Create the plots for nucleotide-level coverage data per sample.
self.plot_TS()
# generate plots for coverage consistency information for each gene.
self.gen_gene_consistency_plots()
# create the gene_class_df
self.init_gene_class_df()
def gen_gene_consistency_plots(self):
""" generate and save the gene consistency plots for each gene."""
if not self.gene_coverage_consistency_dict_initiated:
self.init_gene_coverage_consistency_information()
num_genes, counter = len(
self.profile_db.gene_level_coverage_stats_dict.keys()), 1
progress.new('Plotting gene consistency information')
progress.update('...')
for gene_id in self.profile_db.gene_level_coverage_stats_dict.keys():
if num_genes > 100 and counter % 100 == 0:
self.progress.update('%d of %d genes...' %
(counter, num_genes))
p = MCGPlots(self, gene_id, run=run, progress=progress)
p.plot()
progress.end()
def get_coverage_and_detection_dict(self, bin_id):
_bin = summarizer.Bin(self.summary, bin_id)
self.coverage_values_per_nt = get_coverage_values_per_nucleotide(
_bin.split_coverage_values_per_nt_dict, self.samples)
# getting the total length of all contigs
self.total_length = _bin.total_length
self.init_coverage_and_detection_dataframes(_bin.gene_coverages,
_bin.gene_detection)
def classify(self):
if self.collection_name:
bin_names_in_collection = self.summary.bin_ids
if self.bin_ids_file_path:
filesnpaths.is_file_exists(self.bin_ids_file_path)
bin_names_of_interest = [
line.strip()
for line in open(self.bin_ids_file_path).readlines()
]
missing_bins = [
b for b in bin_names_of_interest
if b not in bin_names_in_collection
]
if len(missing_bins):
raise ConfigError(
"Some bin names you declared do not appear to be in the collection %s. \
These are the bins that are missing: %s, these are the bins that are \
actually in your collection: %s" %
(self.collection_name, missing_bins,
bin_names_in_collection))
elif self.bin_id:
if self.bin_id not in bin_names_in_collection:
raise ConfigError("The bin you declared, %s, does not appear to be in the collection %s." \
% (self.bin_id, self.collection_name))
bin_names_of_interest = [self.bin_id]
else:
bin_names_of_interest = bin_names_in_collection
for bin_id in bin_names_of_interest:
self.run.info_single('Classifying genes in bin: %s' % bin_id)
self.get_coverage_and_detection_dict(bin_id)
self.additional_description = bin_id
self.get_gene_classes()
#self.save_gene_class_information_in_additional_layers(bin_id)
#self.save_samples_information(bin_id)
else:
# No collection provided so running on the entire detection table
self.get_gene_classes()
