def run(self, network, antecedents, out_attributes, user_options,
num_cores, outfile):
from genomicode import genesetlib
import plot_sample_pca
data_node, classify_node = antecedents
result_data = genesetlib.read_tdf(classify_node.identifier,
preserve_spaces=True,
allow_duplicates=True)
for i in result_data:
if i[0] == 'Predicted_class':
legend = i[2]
colors = ['r', 'b', 'g', 'y']
legend_dict = {}
for index, item in enumerate(legend):
if item not in legend_dict:
legend_dict[item] = [index]
else:
legend_dict[item].append(index)
color = [''] * len(legend)
for index, key in enumerate(legend_dict.keys()):
c = colors[index]
for i in legend_dict[key]:
color[i] = c
plot_sample_pca.plot_pca(data_node.identifier, outfile, color, legend)
def read_annotation_descriptor(annotation_descriptor, annotations):
# Return list of (header, annots), where the annots are aligned to
# the annotations.
from genomicode import jmath
from genomicode import genesetlib
filename = annotation_descriptor
assert os.path.exists(filename), "I could not find file: %s" % filename
header2annots = [] # list of (header, annots)
for x in genesetlib.read_tdf(
filename, preserve_spaces=True, allow_duplicates=True):
name, description, annots = x
header2annots.append((name, annots))
assert header2annots, "No annots."
# Find the column that contains the annotations.
header2counts = {}
for (header, annots) in header2annots:
count = len(set(annotations).intersection(annots))
header2counts[header] = count
best_header = best_count = None
for (header, count) in header2counts.iteritems():
if best_count is None or count > best_count:
best_header, best_count = header, count
assert best_count >= len(annotations)/2.0, \
"I could not find the annotations in the descriptor file."
annot_header = best_header
# Align the annotation matrix to the annotations.
annot_annots = None
for (header, annots) in header2annots:
if header == annot_header:
annot_annots = annots
assert annot_annots
I = jmath.match(annotations, annot_annots)
header2annots_aligned = []
for header, annots in header2annots:
annots_aligned = []
for i in I:
if i is None:
annots_aligned.append("")
else:
annots_aligned.append(annots[i])
x = header, annots_aligned
header2annots_aligned.append(x)
header2annots = header2annots_aligned
return header2annots
def read(filename, is_csv=False, header_char=None, nrows=None):
# Everything are strings. No numeric conversion.
import re
from genomicode import genesetlib
delimiter = "\t"
if is_csv:
delimiter = ","
# re.sub takes a lot of time (25% of all running time!). Compile
# it.
re_naive = re.compile("na\\W+ve")
all_headers, all_annots = [], []
all_comments = []
for x in genesetlib.read_tdf(filename,
preserve_spaces=True,
allow_duplicates=True,
delimiter=delimiter,
yield_lines_startswith=header_char,
nrows=nrows):
if type(x) is type(""):
all_comments.append(x)
continue
name, description, annots = x
# Hack: Some files contain special characters, which mess up
# alignment. Fix this here.
# na\xc3\xafve-WIBR3.5 hESC
# na\xe2\x80\x9a\xc3\xa0\xc3\xb6\xe2\x88\x9a\xc3\xb2ve-C1.2 hiPSC
#annots = [re.sub("na\\W+ve", "naive", x) for x in annots]
# This takes a long time. Don't do it unless necessary.
if False:
annots = [re_naive.sub("naive", x) for x in annots]
all_headers.append(name)
all_annots.append(annots)
assert all_headers, "Empty file: %s" % filename
headers_h = uniquify_headers(all_headers)
header2annots = {}
for (header_h, annots) in zip(headers_h, all_annots):
header2annots[header_h] = annots
return AnnotationMatrix(all_headers,
headers_h,
header2annots,
headerlines=all_comments)
def read_gene_descriptor(gene_descriptor, geneset):
# Read pretty names for the genes. gene_descriptor is in the
# format of <filename>,<header>. Return a dictionary of gene ->
# pretty name.
from genomicode import genesetlib
x = gene_descriptor.split(",")
assert len(x) >= 2
filename, pretty_header = x
assert os.path.exists(filename), "I could not find file: %s" % filename
header2genes = {}
for x in genesetlib.read_tdf(
filename, preserve_spaces=True, allow_duplicates=True):
name, description, genes = x
header2genes[name] = genes
assert header2genes, "No genes."
# Find the column that contains the genes in the gene set. Since
# some of the genes may not be annotated, provide some leeway
# here.
header2counts = {}
for (header, genes) in header2genes.iteritems():
count = len(set(geneset).intersection(genes))
header2counts[header] = count
best_header = best_count = None
for (header, count) in header2counts.iteritems():
if best_count is None or count > best_count:
best_header, best_count = header, count
assert best_count >= len(geneset)/2.0, \
"I could not find the genes in the descriptor file."
gene_header = best_header
genes = header2genes[gene_header]
pretty = header2genes[pretty_header]
assert len(genes) == len(pretty)
gene2pretty = {}
for g, p in zip(genes, pretty):
if not g or not p:
continue
gene2pretty[g] = p
return gene2pretty
def read_infile(filename):
from genomicode import genesetlib
name_order = []
name2annots = {}
num_annots = None
for x in genesetlib.read_tdf(filename,
preserve_spaces=True,
allow_duplicates=True):
name, description, annots = x
if num_annots is None:
num_annots = len(annots)
assert len(annots) == num_annots
name_order.append(name)
name2annots[name] = annots
return AnnotationMatrix(name2annots, name_order)
def read_clinical_annotations(M, filename):
# Return a tuple of (Matrix, clinical annotations). The
# annotations are a dictionary of name -> list of values. They
# are aligned with the matrix.
from genomicode import genesetlib
clinical_annots = {}
for x in genesetlib.read_tdf(filename,
preserve_spaces=True,
allow_duplicates=True):
name, description, values = x
clinical_annots[name] = values
# Align the gene scores with the clinical annotations.
x = align_matrix_with_clinical_data(M, clinical_annots)
M, clinical_annots = x
return M, clinical_annots