def test_predict_random_neighbor(self):
"""predict_random_neighbor predicts randomly"""
traits = self.SimpleTreeTraits
tree = self.SimpleTree
result_tree = assign_traits_to_tree(traits,tree)
#If there is only one other valid result, this
#should always be predicted
#self.SimpleTreeTraits =\
# {"A":[1.0,1.0],"E":[1.0,1.0],"F":[0.0,1.0],"D":[0.0,0.0]}
#If self predictions are disallowed, then the prediction for A should
#always come from node D, and be 0,0.
results = predict_random_neighbor(tree,['A'],\
trait_label = "Reconstruction",\
use_self_in_prediction=False)
self.assertEqual(results['A'],[0.0,0.0])
#If use_self is True, ~50% of predictions should be [1.0,1.0] and
# half should be [0.0,0.0]
#Pick repeatedly and make sure frequencies are
#reasonable. The technique is fast, so
#many iterations are reasonable.
iterations = 100000
a_predictions = 0
d_predictions = 0
for i in range(iterations):
results = predict_random_neighbor(tree,['A'],\
trait_label = "Reconstruction",\
use_self_in_prediction=True)
#print results
if results['A'] == [1.0,1.0]:
#print "A pred"
a_predictions += 1
elif results['A'] == [0.0,0.0]:
#print "D pred"
d_predictions +=1
else:
raise RuntimeError(\
"Bad prediction result: Neither node A nor node D traits used in prediction")
#print "All a predictions:",a_predictions
#print "All d predictions:",d_predictions
ratio = float(a_predictions)/float(iterations)
#print "Ratio:", ratio
self.assertFloatEqual(ratio,0.5,eps=1e-2)
def test_predict_random_neighbor(self):
"""predict_random_neighbor predicts randomly"""
traits = self.SimpleTreeTraits
tree = self.SimpleTree
result_tree = assign_traits_to_tree(traits,tree)
#If there is only one other valid result, this
#should always be predicted
#self.SimpleTreeTraits =\
# {"A":[1.0,1.0],"E":[1.0,1.0],"F":[0.0,1.0],"D":[0.0,0.0]}
#If self predictions are disallowed, then the prediction for A should
#always come from node D, and be 0,0.
results = predict_random_neighbor(tree,['A'],\
trait_label = "Reconstruction",\
use_self_in_prediction=False)
self.assertEqual(results['A'],[0.0,0.0])
#If use_self is True, ~50% of predictions should be [1.0,1.0] and
# half should be [0.0,0.0]
#Pick repeatedly and make sure frequencies are
#reasonable. The technique is fast, so
#many iterations are reasonable.
iterations = 100000
a_predictions = 0
d_predictions = 0
for i in range(iterations):
results = predict_random_neighbor(tree,['A'],\
trait_label = "Reconstruction",\
use_self_in_prediction=True)
#print results
if results['A'] == [1.0,1.0]:
#print "A pred"
a_predictions += 1
elif results['A'] == [0.0,0.0]:
#print "D pred"
d_predictions +=1
else:
raise RuntimeError(\
"Bad prediction result: Neither node A nor node D traits used in prediction")
#print "All a predictions:",a_predictions
#print "All d predictions:",d_predictions
ratio = float(a_predictions)/float(iterations)
#print "Ratio:", ratio
self.assertFloatEqual(ratio,0.5,eps=1e-2)
def main():
option_parser, opts, args =\
parse_command_line_parameters(**script_info)
#if we specify we want NSTI only then we have to calculate it first
if opts.output_accuracy_metrics_only:
opts.calculate_accuracy_metrics = True
if opts.verbose:
print "Loading tree from file:", opts.tree
# Load Tree
#tree = LoadTree(opts.tree)
tree = load_picrust_tree(opts.tree, opts.verbose)
table_headers = []
traits = {}
#load the asr trait table using the previous list of functions to order the arrays
if opts.reconstructed_trait_table:
table_headers,traits =\
update_trait_dict_from_file(opts.reconstructed_trait_table)
#Only load confidence intervals on the reconstruction
#If we actually have ASR values in the analysis
if opts.reconstruction_confidence:
if opts.verbose:
print "Loading ASR confidence data from file:",\
opts.reconstruction_confidence
print "Assuming confidence data is of type:", opts.confidence_format
asr_confidence_output = open(opts.reconstruction_confidence)
asr_min_vals,asr_max_vals, params,column_mapping =\
parse_asr_confidence_output(asr_confidence_output,format=opts.confidence_format)
if 'sigma' in params:
brownian_motion_parameter = params['sigma'][0]
else:
brownian_motion_parameter = None
if opts.verbose:
print "Done. Loaded %i confidence interval values." % (
len(asr_max_vals))
print "Brownian motion parameter:", brownian_motion_parameter
else:
brownian_motion_parameter = None
#load the trait table into a dict with organism names as keys and arrays as functions
table_headers,genome_traits =\
update_trait_dict_from_file(opts.observed_trait_table,table_headers)
#Combine the trait tables overwriting the asr ones if they exist in the genome trait table.
traits.update(genome_traits)
# Specify the attribute where we'll store the reconstructions
trait_label = "Reconstruction"
if opts.verbose:
print "Assigning traits to tree..."
# Decorate tree using the traits
tree = assign_traits_to_tree(traits, tree, trait_label=trait_label)
if opts.reconstruction_confidence:
if opts.verbose:
print "Assigning trait confidence intervals to tree..."
tree = assign_traits_to_tree(asr_min_vals,tree,\
trait_label="lower_bound")
tree = assign_traits_to_tree(asr_max_vals,tree,\
trait_label="upper_bound")
if brownian_motion_parameter is None:
if opts.verbose:
print "No Brownian motion parameters loaded. Inferring these from 95% confidence intervals..."
brownian_motion_parameter = get_brownian_motion_param_from_confidence_intervals(tree,\
upper_bound_trait_label="upper_bound",\
lower_bound_trait_label="lower_bound",\
trait_label=trait_label,\
confidence=0.95)
if opts.verbose:
print "Inferred the following rate parameters:", brownian_motion_parameter
if opts.verbose:
print "Collecting list of nodes to predict..."
#Start by predict all tip nodes.
nodes_to_predict = [tip.Name for tip in tree.tips()]
if opts.verbose:
print "Found %i nodes to predict." % len(nodes_to_predict)
if opts.limit_predictions_to_organisms:
organism_id_str = opts.limit_predictions_to_organisms
ok_organism_ids = organism_id_str.split(',')
ok_organism_ids = [n.strip() for n in ok_organism_ids]
for f in set_label_conversion_fns(True, True):
ok_organism_ids = [f(i) for i in ok_organism_ids]
if opts.verbose:
print "Limiting predictions to user-specified ids:",\
",".join(ok_organism_ids)
if not ok_organism_ids:
raise RuntimeError(\
"Found no valid ids in input: %s. Were comma-separated ids specified on the command line?"\
% opts.limit_predictions_to_organisms)
nodes_to_predict =\
[n for n in nodes_to_predict if n in ok_organism_ids]
if not nodes_to_predict:
raise RuntimeError(\
"Filtering by user-specified ids resulted in an empty set of nodes to predict. Are the ids on the commmand-line and tree ids in the same format? Example tree tip name: %s, example OTU id name: %s" %([tip.Name for tip in tree.tips()][0],ok_organism_ids[0]))
if opts.verbose:
print "After filtering organisms to predict by the ids specified on the commandline, %i nodes remain to be predicted" % (
len(nodes_to_predict))
if opts.limit_predictions_by_otu_table:
if opts.verbose:
print "Limiting predictions to ids in user-specified OTU table:",\
opts.limit_predictions_by_otu_table
otu_table = open(opts.limit_predictions_by_otu_table, "U")
#Parse OTU table for ids
otu_ids =\
extract_ids_from_table(otu_table.readlines(),delimiter="\t")
if not otu_ids:
raise RuntimeError(\
"Found no valid ids in input OTU table: %s. Is the path correct?"\
% opts.limit_predictions_by_otu_table)
nodes_to_predict =\
[n for n in nodes_to_predict if n in otu_ids]
if not nodes_to_predict:
raise RuntimeError(\
"Filtering by OTU table resulted in an empty set of nodes to predict. Are the OTU ids and tree ids in the same format? Example tree tip name: %s, example OTU id name: %s" %([tip.Name for tip in tree.tips()][0],otu_ids[0]))
if opts.verbose:
print "After filtering by OTU table, %i nodes remain to be predicted" % (
len(nodes_to_predict))
# Calculate accuracy of PICRUST for the given tree, sequenced genomes
# and set of ndoes to predict
accuracy_metrics = ['NSTI']
accuracy_metric_results = None
if opts.calculate_accuracy_metrics:
if opts.verbose:
print "Calculating accuracy metrics: %s" % (
[",".join(accuracy_metrics)])
accuracy_metric_results = {}
if 'NSTI' in accuracy_metrics:
nsti_result,min_distances =\
calc_nearest_sequenced_taxon_index(tree,\
limit_to_tips = nodes_to_predict,\
trait_label = trait_label, verbose=opts.verbose)
#accuracy_metric_results['NSTI'] = nsti_result
for organism in min_distances.keys():
accuracy_metric_results[organism] = {
'NSTI': min_distances[organism]
}
if opts.verbose:
print "NSTI:", nsti_result
if opts.output_accuracy_metrics_only:
#Write accuracy metrics to file
if opts.verbose:
print "Writing accuracy metrics to file:", opts.output_accuracy_metrics
f = open(opts.output_accuracy_metrics_only, 'w+')
f.write("metric\torganism\tvalue\n")
lines = []
for organism in accuracy_metric_results.keys():
for metric in accuracy_metric_results[organism].keys():
lines.append('\t'.join([metric,organism,\
str(accuracy_metric_results[organism][metric])])+'\n')
f.writelines(sorted(lines))
f.close()
exit()
if opts.verbose:
print "Generating predictions using method:", opts.prediction_method
if opts.weighting_method == 'exponential':
#For now, use exponential weighting
weight_fn = make_neg_exponential_weight_fn(e)
variances = None #Overwritten by methods that calc variance
confidence_intervals = None #Overwritten by methods that calc variance
if opts.prediction_method == 'asr_and_weighting':
# Perform predictions using reconstructed ancestral states
if opts.reconstruction_confidence:
predictions,variances,confidence_intervals =\
predict_traits_from_ancestors(tree,nodes_to_predict,\
trait_label=trait_label,\
lower_bound_trait_label="lower_bound",\
upper_bound_trait_label="upper_bound",\
calc_confidence_intervals = True,\
brownian_motion_parameter=brownian_motion_parameter,\
weight_fn =weight_fn,verbose=opts.verbose)
else:
predictions =\
predict_traits_from_ancestors(tree,nodes_to_predict,\
trait_label=trait_label,\
weight_fn =weight_fn,verbose=opts.verbose)
elif opts.prediction_method == 'weighting_only':
#Ignore ancestral information
predictions =\
weighted_average_tip_prediction(tree,nodes_to_predict,\
trait_label=trait_label,\
weight_fn =weight_fn,verbose=opts.verbose)
elif opts.prediction_method == 'nearest_neighbor':
predictions = predict_nearest_neighbor(tree,nodes_to_predict,\
trait_label=trait_label,tips_only = True)
elif opts.prediction_method == 'random_neighbor':
predictions = predict_random_neighbor(tree,\
nodes_to_predict,trait_label=trait_label)
if opts.verbose:
print "Done making predictions."
make_output_dir_for_file(opts.output_trait_table)
out_fh = open(opts.output_trait_table, 'w')
#Generate the table of biom predictions
if opts.verbose:
print "Converting results to .biom format for output..."
biom_predictions=biom_table_from_predictions(predictions,table_headers,\
observation_metadata=None,\
sample_metadata=accuracy_metric_results,convert_to_int=False)
if opts.verbose:
print "Writing prediction results to file: ", opts.output_trait_table
if opts.output_precalc_file_in_biom:
#write biom table to file
write_biom_table(biom_predictions, opts.output_trait_table)
else:
#convert to precalc (tab-delimited) format
out_fh = open(opts.output_trait_table, 'w')
out_fh.write(convert_biom_to_precalc(biom_predictions))
out_fh.close()
#Write out variance information to file
if variances:
if opts.verbose:
print "Converting variances to BIOM format"
if opts.output_precalc_file_in_biom:
suffix = '.biom'
else:
suffix = '.tab'
biom_prediction_variances=biom_table_from_predictions({k:v['variance'] for k,v in variances.iteritems()},table_headers,\
observation_metadata=None,\
sample_metadata=None,convert_to_int=False)
outfile_base, extension = splitext(opts.output_trait_table)
variance_outfile = outfile_base + "_variances" + suffix
make_output_dir_for_file(variance_outfile)
if opts.verbose:
print "Writing variance information to file:", variance_outfile
if opts.output_precalc_file_in_biom:
write_biom_table(biom_prediction_variances, variance_outfile)
else:
open(variance_outfile,'w').write(\
convert_biom_to_precalc(biom_prediction_variances))
if confidence_intervals:
if opts.verbose:
print "Converting upper confidence interval values to BIOM format"
biom_prediction_upper_CI=biom_table_from_predictions({k:v['upper_CI'] for k,v in confidence_intervals.iteritems()},table_headers,\
observation_metadata=None,\
sample_metadata=None,convert_to_int=False)
outfile_base, extension = splitext(opts.output_trait_table)
upper_CI_outfile = outfile_base + "_upper_CI" + suffix
make_output_dir_for_file(upper_CI_outfile)
if opts.verbose:
print "Writing upper confidence limit information to file:", upper_CI_outfile
if opts.output_precalc_file_in_biom:
write_biom_table(biom_prediction_upper_CI, upper_CI_outfile)
else:
open(upper_CI_outfile,'w').write(\
convert_biom_to_precalc(biom_prediction_upper_CI))
biom_prediction_lower_CI=biom_table_from_predictions({k:v['lower_CI'] for k,v in confidence_intervals.iteritems()},table_headers,\
observation_metadata=None,\
sample_metadata=None,convert_to_int=False)
outfile_base, extension = splitext(opts.output_trait_table)
lower_CI_outfile = outfile_base + "_lower_CI" + suffix
make_output_dir_for_file(lower_CI_outfile)
if opts.verbose:
print "Writing lower confidence limit information to file", lower_CI_outfile
if opts.output_precalc_file_in_biom:
write_biom_table(biom_prediction_lower_CI, lower_CI_outfile)
else:
open(lower_CI_outfile,'w').write(\
convert_biom_to_precalc(biom_prediction_lower_CI))
def main():
option_parser, opts, args =\
parse_command_line_parameters(**script_info)
#if we specify we want NSTI only then we have to calculate it first
if opts.output_accuracy_metrics_only:
opts.calculate_accuracy_metrics=True
if opts.verbose:
print "Loading tree from file:", opts.tree
if opts.no_round:
round_opt = False
else:
round_opt = True
# Load Tree
tree = load_picrust_tree(opts.tree, opts.verbose)
table_headers=[]
traits={}
#load the asr trait table using the previous list of functions to order the arrays
if opts.reconstructed_trait_table:
table_headers,traits =\
update_trait_dict_from_file(opts.reconstructed_trait_table)
#Only load confidence intervals on the reconstruction
#If we actually have ASR values in the analysis
if opts.reconstruction_confidence:
if opts.verbose:
print "Loading ASR confidence data from file:",\
opts.reconstruction_confidence
print "Assuming confidence data is of type:",opts.confidence_format
asr_confidence_output = open(opts.reconstruction_confidence)
asr_min_vals,asr_max_vals, params,column_mapping =\
parse_asr_confidence_output(asr_confidence_output,format=opts.confidence_format)
if 'sigma' in params:
brownian_motion_parameter = params['sigma'][0]
else:
brownian_motion_parameter = None
if opts.verbose:
print "Done. Loaded %i confidence interval values." %(len(asr_max_vals))
print "Brownian motion parameter:",brownian_motion_parameter
else:
brownian_motion_parameter = None
#load the trait table into a dict with organism names as keys and arrays as functions
table_headers,genome_traits =\
update_trait_dict_from_file(opts.observed_trait_table,table_headers)
#Combine the trait tables overwriting the asr ones if they exist in the genome trait table.
traits.update(genome_traits)
# Specify the attribute where we'll store the reconstructions
trait_label = "Reconstruction"
if opts.verbose:
print "Assigning traits to tree..."
# Decorate tree using the traits
tree = assign_traits_to_tree(traits,tree, trait_label=trait_label)
if opts.reconstruction_confidence:
if opts.verbose:
print "Assigning trait confidence intervals to tree..."
tree = assign_traits_to_tree(asr_min_vals,tree,\
trait_label="lower_bound")
tree = assign_traits_to_tree(asr_max_vals,tree,\
trait_label="upper_bound")
if brownian_motion_parameter is None:
if opts.verbose:
print "No Brownian motion parameters loaded. Inferring these from 95% confidence intervals..."
brownian_motion_parameter = get_brownian_motion_param_from_confidence_intervals(tree,\
upper_bound_trait_label="upper_bound",\
lower_bound_trait_label="lower_bound",\
trait_label=trait_label,\
confidence=0.95)
if opts.verbose:
print "Inferred the following rate parameters:",brownian_motion_parameter
if opts.verbose:
print "Collecting list of nodes to predict..."
#Start by predict all tip nodes.
nodes_to_predict = [tip.Name for tip in tree.tips()]
if opts.verbose:
print "Found %i nodes to predict." % len(nodes_to_predict)
if opts.limit_predictions_to_organisms:
organism_id_str = opts.limit_predictions_to_organisms
ok_organism_ids = organism_id_str.split(',')
ok_organism_ids = [n.strip() for n in ok_organism_ids]
for f in set_label_conversion_fns(True,True):
ok_organism_ids = [f(i) for i in ok_organism_ids]
if opts.verbose:
print "Limiting predictions to user-specified ids:",\
",".join(ok_organism_ids)
if not ok_organism_ids:
raise RuntimeError(\
"Found no valid ids in input: %s. Were comma-separated ids specified on the command line?"\
% opts.limit_predictions_to_organisms)
nodes_to_predict =\
[n for n in nodes_to_predict if n in ok_organism_ids]
if not nodes_to_predict:
raise RuntimeError(\
"Filtering by user-specified ids resulted in an empty set of nodes to predict. Are the ids on the commmand-line and tree ids in the same format? Example tree tip name: %s, example OTU id name: %s" %([tip.Name for tip in tree.tips()][0],ok_organism_ids[0]))
if opts.verbose:
print "After filtering organisms to predict by the ids specified on the commandline, %i nodes remain to be predicted" %(len(nodes_to_predict))
if opts.limit_predictions_by_otu_table:
if opts.verbose:
print "Limiting predictions to ids in user-specified OTU table:",\
opts.limit_predictions_by_otu_table
otu_table = open(opts.limit_predictions_by_otu_table,"U")
#Parse OTU table for ids
otu_ids =\
extract_ids_from_table(otu_table.readlines(),delimiter="\t")
if not otu_ids:
raise RuntimeError(\
"Found no valid ids in input OTU table: %s. Is the path correct?"\
% opts.limit_predictions_by_otu_table)
nodes_to_predict =\
[n for n in nodes_to_predict if n in otu_ids]
if not nodes_to_predict:
raise RuntimeError(\
"Filtering by OTU table resulted in an empty set of nodes to predict. Are the OTU ids and tree ids in the same format? Example tree tip name: %s, example OTU id name: %s" %([tip.Name for tip in tree.tips()][0],otu_ids[0]))
if opts.verbose:
print "After filtering by OTU table, %i nodes remain to be predicted" %(len(nodes_to_predict))
# Calculate accuracy of PICRUST for the given tree, sequenced genomes
# and set of ndoes to predict
accuracy_metrics = ['NSTI']
accuracy_metric_results = None
if opts.calculate_accuracy_metrics:
if opts.verbose:
print "Calculating accuracy metrics: %s" %([",".join(accuracy_metrics)])
accuracy_metric_results = {}
if 'NSTI' in accuracy_metrics:
nsti_result,min_distances =\
calc_nearest_sequenced_taxon_index(tree,\
limit_to_tips = nodes_to_predict,\
trait_label = trait_label, verbose=opts.verbose)
#accuracy_metric_results['NSTI'] = nsti_result
for organism in min_distances.keys():
accuracy_metric_results[organism] = {'NSTI': min_distances[organism]}
if opts.verbose:
print "NSTI:", nsti_result
if opts.output_accuracy_metrics_only:
#Write accuracy metrics to file
if opts.verbose:
print "Writing accuracy metrics to file:",opts.output_accuracy_metrics
f = open(opts.output_accuracy_metrics_only,'w+')
f.write("metric\torganism\tvalue\n")
lines =[]
for organism in accuracy_metric_results.keys():
for metric in accuracy_metric_results[organism].keys():
lines.append('\t'.join([metric,organism,\
str(accuracy_metric_results[organism][metric])])+'\n')
f.writelines(sorted(lines))
f.close()
exit()
if opts.verbose:
print "Generating predictions using method:",opts.prediction_method
if opts.weighting_method == 'exponential':
#For now, use exponential weighting
weight_fn = make_neg_exponential_weight_fn(e)
variances=None #Overwritten by methods that calc variance
confidence_intervals=None #Overwritten by methods that calc variance
if opts.prediction_method == 'asr_and_weighting':
# Perform predictions using reconstructed ancestral states
if opts.reconstruction_confidence:
predictions,variances,confidence_intervals =\
predict_traits_from_ancestors(tree,nodes_to_predict,\
trait_label=trait_label,\
lower_bound_trait_label="lower_bound",\
upper_bound_trait_label="upper_bound",\
calc_confidence_intervals = True,\
brownian_motion_parameter=brownian_motion_parameter,\
weight_fn=weight_fn,verbose=opts.verbose,
round_predictions=round_opt)
else:
predictions =\
predict_traits_from_ancestors(tree,nodes_to_predict,\
trait_label=trait_label,\
weight_fn =weight_fn,verbose=opts.verbose,
round_predictions=round_opt)
elif opts.prediction_method == 'weighting_only':
#Ignore ancestral information
predictions =\
weighted_average_tip_prediction(tree,nodes_to_predict,\
trait_label=trait_label,\
weight_fn =weight_fn,verbose=opts.verbose)
elif opts.prediction_method == 'nearest_neighbor':
predictions = predict_nearest_neighbor(tree,nodes_to_predict,\
trait_label=trait_label,tips_only = True)
elif opts.prediction_method == 'random_neighbor':
predictions = predict_random_neighbor(tree,\
nodes_to_predict,trait_label=trait_label)
if opts.verbose:
print "Done making predictions."
make_output_dir_for_file(opts.output_trait_table)
out_fh=open(opts.output_trait_table,'w')
#Generate the table of biom predictions
if opts.verbose:
print "Converting results to .biom format for output..."
biom_predictions=biom_table_from_predictions(predictions,table_headers,\
observation_metadata=None,\
sample_metadata=accuracy_metric_results,convert_to_int=False)
if opts.verbose:
print "Writing prediction results to file: ",opts.output_trait_table
if opts.output_precalc_file_in_biom:
#write biom table to file
write_biom_table(biom_predictions, opts.output_trait_table)
else:
#convert to precalc (tab-delimited) format
out_fh = open(opts.output_trait_table, 'w')
out_fh.write(convert_biom_to_precalc(biom_predictions))
out_fh.close()
#Write out variance information to file
if variances:
if opts.verbose:
print "Converting variances to BIOM format"
if opts.output_precalc_file_in_biom:
suffix='.biom'
else:
suffix='.tab'
biom_prediction_variances=biom_table_from_predictions({k:v['variance'] for k,v in variances.iteritems()},table_headers,\
observation_metadata=None,\
sample_metadata=None,convert_to_int=False)
outfile_base,extension = splitext(opts.output_trait_table)
variance_outfile = outfile_base+"_variances"+suffix
make_output_dir_for_file(variance_outfile)
if opts.verbose:
print "Writing variance information to file:",variance_outfile
if opts.output_precalc_file_in_biom:
write_biom_table(biom_prediction_variances, variance_outfile)
else:
open(variance_outfile,'w').write(\
convert_biom_to_precalc(biom_prediction_variances))
if confidence_intervals:
if opts.verbose:
print "Converting upper confidence interval values to BIOM format"
biom_prediction_upper_CI=biom_table_from_predictions({k:v['upper_CI'] for k,v in confidence_intervals.iteritems()},table_headers,\
observation_metadata=None,\
sample_metadata=None,convert_to_int=False)
outfile_base,extension = splitext(opts.output_trait_table)
upper_CI_outfile = outfile_base+"_upper_CI"+suffix
make_output_dir_for_file(upper_CI_outfile)
if opts.verbose:
print "Writing upper confidence limit information to file:",upper_CI_outfile
if opts.output_precalc_file_in_biom:
write_biom_table(biom_prediction_upper_CI, upper_CI_outfile)
else:
open(upper_CI_outfile,'w').write(\
convert_biom_to_precalc(biom_prediction_upper_CI))
biom_prediction_lower_CI=biom_table_from_predictions({k:v['lower_CI'] for k,v in confidence_intervals.iteritems()},table_headers,\
observation_metadata=None,\
sample_metadata=None,convert_to_int=False)
outfile_base,extension = splitext(opts.output_trait_table)
lower_CI_outfile = outfile_base+"_lower_CI"+suffix
make_output_dir_for_file(lower_CI_outfile)
if opts.verbose:
print "Writing lower confidence limit information to file",lower_CI_outfile
if opts.output_precalc_file_in_biom:
write_biom_table(biom_prediction_lower_CI, lower_CI_outfile)
else:
open(lower_CI_outfile,'w').write(\
convert_biom_to_precalc(biom_prediction_lower_CI))
def main():
option_parser, opts, args =\
parse_command_line_parameters(**script_info)
if opts.verbose:
print "Loading tree from file:", opts.tree
# Load Tree
#tree = LoadTree(opts.tree)
tree = load_picrust_tree(opts.tree, opts.verbose)
table_headers =[]
traits={}
#load the asr trait table using the previous list of functions to order the arrays
if opts.reconstructed_trait_table:
table_headers,traits =\
update_trait_dict_from_file(opts.reconstructed_trait_table)
#Only load confidence intervals on the reconstruction
#If we actually have ASR values in the analysis
if opts.reconstruction_confidence:
if opts.verbose:
print "Loading ASR confidence data from file:",\
opts.reconstruction_confidence
asr_confidence_output = open(opts.reconstruction_confidence)
asr_min_vals,asr_max_vals, params,column_mapping =\
parse_asr_confidence_output(asr_confidence_output)
brownian_motion_parameter = params['sigma'][0]
brownian_motion_error = params['sigma'][1]
if opts.verbose:
print "Done. Loaded %i confidence interval values." %(len(asr_max_vals))
print "Brownian motion parameter:",brownian_motion_parameter
else:
brownian_motion_parameter = None
#load the trait table into a dict with organism names as keys and arrays as functions
table_headers,genome_traits =\
update_trait_dict_from_file(opts.observed_trait_table,table_headers)
#Combine the trait tables overwriting the asr ones if they exist in the genome trait table.
traits.update(genome_traits)
# Specify the attribute where we'll store the reconstructions
trait_label = "Reconstruction"
if opts.verbose:
print "Assigning traits to tree..."
# Decorate tree using the traits
tree = assign_traits_to_tree(traits,tree, trait_label=trait_label)
if opts.reconstruction_confidence:
if opts.verbose:
print "Assigning trait confidence intervals to tree..."
tree = assign_traits_to_tree(asr_min_vals,tree,\
trait_label="lower_bound")
tree = assign_traits_to_tree(asr_max_vals,tree,\
trait_label="upper_bound")
if opts.verbose:
print "Collecting list of nodes to predict..."
#Start by predict all tip nodes.
nodes_to_predict = [tip.Name for tip in tree.tips()]
if opts.verbose:
print "Found %i nodes to predict." % len(nodes_to_predict)
if opts.limit_predictions_to_organisms:
organism_id_str = opts.limit_predictions_to_organisms
ok_organism_ids = organism_id_str.split(',')
ok_organism_ids = [n.strip() for n in ok_organism_ids]
for f in set_label_conversion_fns(True,True):
ok_organism_ids = [f(i) for i in ok_organism_ids]
if opts.verbose:
print "Limiting predictions to user-specified ids:",\
",".join(ok_organism_ids)
if not ok_organism_ids:
raise RuntimeError(\
"Found no valid ids in input: %s. Were comma-separated ids specified on the command line?"\
% opts.limit_predictions_to_organisms)
nodes_to_predict =\
[n for n in nodes_to_predict if n in ok_organism_ids]
if not nodes_to_predict:
raise RuntimeError(\
"Filtering by user-specified ids resulted in an empty set of nodes to predict. Are the ids on the commmand-line and tree ids in the same format? Example tree tip name: %s, example OTU id name: %s" %([tip.Name for tip in tree.tips()][0],ok_organism_ids[0]))
if opts.verbose:
print "After filtering organisms to predict by the ids specified on the commandline, %i nodes remain to be predicted" %(len(nodes_to_predict))
if opts.limit_predictions_by_otu_table:
if opts.verbose:
print "Limiting predictions to ids in user-specified OTU table:",\
opts.limit_predictions_by_otu_table
otu_table = open(opts.limit_predictions_by_otu_table,"U")
#Parse OTU table for ids
otu_ids =\
extract_ids_from_table(otu_table.readlines(),delimiter="\t")
if not otu_ids:
raise RuntimeError(\
"Found no valid ids in input OTU table: %s. Is the path correct?"\
% opts.limit_predictions_by_otu_table)
nodes_to_predict =\
[n for n in nodes_to_predict if n in otu_ids]
if not nodes_to_predict:
raise RuntimeError(\
"Filtering by OTU table resulted in an empty set of nodes to predict. Are the OTU ids and tree ids in the same format? Example tree tip name: %s, example OTU id name: %s" %([tip.Name for tip in tree.tips()][0],otu_ids[0]))
if opts.verbose:
print "After filtering by OTU table, %i nodes remain to be predicted" %(len(nodes_to_predict))
# Calculate accuracy of PICRUST for the given tree, sequenced genomes
# and set of ndoes to predict
accuracy_metrics = ['NSTI']
accuracy_metric_results = None
if opts.output_accuracy_metrics:
if opts.verbose:
print "Calculating accuracy metrics: %s" %([",".join(accuracy_metrics)])
accuracy_metric_results = {}
if 'NSTI' in accuracy_metrics:
nsti_result,min_distances =\
calc_nearest_sequenced_taxon_index(tree,\
limit_to_tips = nodes_to_predict,\
trait_label = trait_label, verbose=opts.verbose)
#accuracy_metric_results['NSTI'] = nsti_result
for organism in min_distances.keys():
accuracy_metric_results[organism] = {'NSTI': min_distances[organism]}
if opts.verbose:
print "NSTI:", nsti_result
#Write accuracy metrics to file
if opts.verbose:
print "Writing accuracy metrics to file:",opts.output_accuracy_metrics
f = open(opts.output_accuracy_metrics,'w+')
lines = ["metric\torganism\tvalue\n"]
for organism in accuracy_metric_results.keys():
for metric in accuracy_metric_results[organism].keys():
lines.append('\t'.join([metric,organism,\
str(accuracy_metric_results[organism][metric])])+'\n')
f.writelines(sorted(lines))
f.close()
if opts.verbose:
print "Generating predictions using method:",opts.prediction_method
if opts.weighting_method == 'exponential':
#For now, use exponential weighting
weight_fn = make_neg_exponential_weight_fn(e)
elif opts.weighting_method == 'linear':
#Linear weight function
weight_fn = linear_weight
elif opts.weighting_method == 'equal_weight':
weight_fn = equal_weight
variances=None #Overwritten by methods that calc variance
if opts.prediction_method == 'asr_and_weighting':
if opts.reconstruction_confidence:
# Perform predictions using reconstructed ancestral states
predictions,variances =\
predict_traits_from_ancestors(tree,nodes_to_predict,\
trait_label=trait_label,\
lower_bound_trait_label="lower_bound",\
upper_bound_trait_label="upper_bound",\
calc_confidence_intervals = True,\
brownian_motion_parameter=brownian_motion_parameter,\
use_self_in_prediction = True,\
weight_fn =weight_fn,verbose=opts.verbose)
else:
predictions =\
predict_traits_from_ancestors(tree,nodes_to_predict,\
trait_label=trait_label,\
use_self_in_prediction = True,\
weight_fn =weight_fn,verbose=opts.verbose)
elif opts.prediction_method == 'weighting_only':
#Ignore ancestral information
predictions =\
weighted_average_tip_prediction(tree,nodes_to_predict,\
trait_label=trait_label,\
use_self_in_prediction = True,\
weight_fn =weight_fn,verbose=opts.verbose)
elif opts.prediction_method == 'nearest_neighbor':
predictions = predict_nearest_neighbor(tree,nodes_to_predict,\
trait_label=trait_label,\
use_self_in_prediction = True, tips_only = True)
elif opts.prediction_method == 'random_neighbor':
predictions = predict_random_neighbor(tree,\
nodes_to_predict,trait_label=trait_label,\
use_self_in_prediction = True)
else:
error_template =\
"Prediction method '%s' is not supported. Valid methods are: %s'"
error_text = error_template %(opts.prediction_method,\
", ".join(METHOD_CHOICES))
if opts.verbose:
print "Converting results to .biom format for output..."
#convert to biom format (and transpose)
biom_predictions=biom_table_from_predictions(predictions,table_headers)
#In the .biom table, organisms are 'samples' and traits are 'observations
#(by analogy with a metagenomic sample)
#Therefore, we associate the trait variances with the per-observation metadata
#print "variances:",variances
#print "BIOM observations:", [o for o in biom_predictions.iterObservations()]
#print "BIOM samples:", [s for s in biom_predictions.iterSamples()]
if variances is not None:
if opts.verbose:
print "Adding variance information to output .biom table, as per-observation metadata with key 'variance'..."
biom_predictions.addSampleMetadata(variances)
if accuracy_metric_results is not None:
if opts.verbose:
print "Adding accuracy metrics (%s) to biom table as per-observation metadata..." %(",".join(accuracy_metrics))
biom_predictions.addSampleMetadata(accuracy_metric_results)
#Add variance information as per observation metadata
if opts.verbose:
print "Writing biom format prediction results to file: ",opts.output_trait_table
#write biom table to file
make_output_dir_for_file(opts.output_trait_table)
open(opts.output_trait_table,'w').write(\
format_biom_table(biom_predictions))