def __init__(self, args, run=terminal.Run(), progress=terminal.Progress()):
self.run = run
self.progress = progress
A = lambda x: (args.__dict__[x]
if x in args.__dict__ else None) if args else None
if self.mode == 'train':
self.genomes_dir = os.path.abspath(A('genomes_dir'))
self.classifier_output_path = os.path.abspath(A('output'))
if A('classifier'):
raise ConfigError(
"You should not initialize the domain training class with a input classifier path (`args.classifier`)."
)
if not self.genomes_dir:
raise ConfigError(
"You must provide a genomes directory. Please read the help menu if you are not sure\
how the contents of this directory should look like."
)
filesnpaths.is_output_file_writable(self.classifier_output_path)
filesnpaths.is_file_exists(self.genomes_dir)
elif self.mode == 'predict':
if A('output'):
raise ConfigError(
"You should not initialize the domain prediction class with an output classifier path (`args.output`)."
)
default_classifier_path = 'misc/SCGDOMAINCLASSIFIER.rf'
self.input_classifier_path = A('classifier') or os.path.join(
os.path.dirname(anvio.data.__file__), default_classifier_path)
if A('classifier'):
filesnpaths.is_file_exists(self.input_classifier_path)
else:
if not filesnpaths.is_file_exists(self.input_classifier_path,
dont_raise=True):
raise ConfigError(
"Somehow, this anvi'o installation dose not seem to have a SCG domain classifier. This is one of\
those anvi'o things that should never happen. If you are an anvi'o user, please feel free to panic :(\
If you are an anvi'o developer, what you need to do is to follow the instructions in \
`anvi-script-gen-scg-domain-classifier` with a reasonable set of genomes and store the resulting\
classifier at the default anvi'o path of /blah/blah/anvio/data/%s."
% (default_classifier_path))
self.rf = RF(self.input_classifier_path,
r=self.run,
p=self.progress)
self.rf.initialize_classifier()
else:
raise ConfigError(
"Someone initialized the SCG domain classifier class without an explicit mode :("
)
self.SCG_sources = [
d for d in hmm_data.sources
if hmm_data.sources[d]['kind'] == 'singlecopy'
]
self.SCG_domains = sorted([
hmm_data.sources[source]['domain'] for source in self.SCG_sources
])
self.SCG_domain_to_source = dict([(hmm_data.sources[source]['domain'],
source)
for source in self.SCG_sources])
if not len(self.SCG_sources):
raise ConfigError(
"There is something wrong :( There is not even a single SCG source found. Usually\
anvi'o comes with multiple of them :/")
if len(self.SCG_sources) == 1:
raise ConfigError(
"There is only a single SCG source in your anvi'o installation. It is OK if you are\
being a hacker and playing with things, but there is no logic behind creating a\
classifier with a single class.")
if len(self.SCG_domains) != len(set(self.SCG_domains)):
raise ConfigError(
"Something is wrong. For each domain, there must be a single sinlge-copy core gene\
source.")
self.data, self.labels, self.features = [], [], []
for domain in self.SCG_domains:
self.features.extend(
sorted(hmm_data.sources[self.SCG_domain_to_source[domain]]
['genes']))
self.run.info('SCG domain classifier mode', self.mode)
self.run.info("SCG domains found", ', '.join(self.SCG_domains))
self.run.info("Num features", len(self.features))
def __init__(self, args, run=terminal.Run(), progress=terminal.Progress()):
self.run = run
self.progress = progress
A = lambda x: (args.__dict__[x] if x in args.__dict__ else None) if args else None
if self.mode == 'train':
self.genomes_dir = os.path.abspath(A('genomes_dir'))
self.classifier_output_path = os.path.abspath(A('output'))
if A('classifier'):
raise ConfigError("You should not initialize the domain training class with a input classifier path (`args.classifier`).")
if not self.genomes_dir:
raise ConfigError("You must provide a genomes directory. Please read the help menu if you are not sure\
how the contents of this directory should look like.")
filesnpaths.is_output_file_writable(self.classifier_output_path)
filesnpaths.is_file_exists(self.genomes_dir)
elif self.mode == 'predict':
if A('output'):
raise ConfigError("You should not initialize the domain prediction class with an output classifier path (`args.output`).")
default_classifier_path = 'misc/SCGDOMAINCLASSIFIER.rf'
self.input_classifier_path = A('classifier') or os.path.join(os.path.dirname(anvio.data.__file__), default_classifier_path)
if A('classifier'):
filesnpaths.is_file_exists(self.input_classifier_path)
else:
if not filesnpaths.is_file_exists(self.input_classifier_path, dont_raise=True):
raise ConfigError("Somehow, this anvi'o installation dose not seem to have a SCG domain classifier. This is one of\
those anvi'o things that should never happen. If you are an anvi'o user, please feel free to panic :(\
If you are an anvi'o developer, what you need to do is to follow the instructions in \
`anvi-script-gen-scg-domain-classifier` with a reasonable set of genomes and store the resulting\
classifier at the default anvi'o path of /blah/blah/anvio/data/%s." % (default_classifier_path))
self.rf = RF(self.input_classifier_path, r=self.run, p=self.progress)
self.rf.initialize_classifier()
else:
raise ConfigError("Someone initialized the SCG domain classifier class without an explicit mode :(")
self.SCG_sources = [d for d in hmm_data.sources if hmm_data.sources[d]['kind'] == 'singlecopy']
self.SCG_domains = sorted([hmm_data.sources[source]['domain'] for source in self.SCG_sources])
self.SCG_domain_to_source = dict([(hmm_data.sources[source]['domain'], source) for source in self.SCG_sources])
if not len(self.SCG_sources):
raise ConfigError("There is something wrong :( There is not even a single SCG source found. Usually\
anvi'o comes with multiple of them :/")
if len(self.SCG_sources) == 1:
raise ConfigError("There is only a single SCG source in your anvi'o installation. It is OK if you are\
being a hacker and playing with things, but there is no logic behind creating a\
classifier with a single class.")
if len(self.SCG_domains) != len(set(self.SCG_domains)):
raise ConfigError("Something is wrong. For each domain, there must be a single sinlge-copy core gene\
source.")
self.data, self.labels, self.features = [], [], []
for domain in self.SCG_domains:
self.features.extend(sorted(hmm_data.sources[self.SCG_domain_to_source[domain]]['genes']))
self.run.info('SCG domain classifier mode', self.mode)
self.run.info("SCG domains found", ', '.join(self.SCG_domains))
self.run.info("Num features", len(self.features))
def train(self):
rf = RF(self.classifier_output_path, r=self.run, p=self.progress)
rf.train(self.features, self.data, self.labels)
def train(self):
rf = RF(self.classifier_output_path, r=self.run, p=self.progress)
rf.train(self.features, self.data, self.labels)