def __call__(self, content):
from subprocess import Popen, PIPE
from sys import stdout
if not isinstance(content, (str, bytes)):
raise TypeError('Unknown input type for paging.')
try:
cmd = ['less', '-F', '-R', '-S', '-X', '-K']
proc = Popen(cmd, stdin=PIPE, stdout=stdout)
if isinstance(content, str):
proc._stdin_write(content.encode())
else:
proc._stdin_write(content)
proc.stdin.close()
proc.wait()
except KeyboardInterrupt:
pass
def gsva(
expression: Union[ExpressionWithControls, Profile], gene_sets_path: str, method: str = 'gsva',
single_sample=False, permutations=1000, mx_diff=True, cores=1, _cache=True, limit_to_gene_sets=False,
verbose=False
):
"""
Excerpt from GSVA documentation:
An important argument of the gsva() function is the flag mx.diff which is set to TRUE by default.
Under this default setting, GSVA enrichment scores are calculated using Equation 5, and therefore, are
more amenable by analysis techniques that assume the data to be normally distributed. When setting
mx.diff=FALSE , then Equation 4 is employed, calculating enrichment in an analogous way to classical
GSEA which typically provides a bimodal distribution of GSVA enrichment scores for each gene.
"""
if not single_sample and permutations:
raise warn('permutations are not supported when not single_sample')
key = (expression.hashable, method, gene_sets_path)
if key in GSVA_CACHE:
return GSVA_CACHE[key]
if single_sample:
assert isinstance(expression, Profile)
joined = DataFrame(
concat([expression.top.up, expression.top.down]),
)
joined.columns = ['condition']
joined['control'] = 0
joined.index = joined.index.astype(str)
expression_classes = Series(['case', 'control'])
expression = joined
else:
joined = expression.joined
joined = DataFrame(joined)
joined.index = joined.index.astype(str)
nulls = joined.isnull().any(axis=0).reset_index(drop=True)
if nulls.any():
print(f'Following columns contain nulls and will be skipped: {list(joined.columns[nulls])}')
joined = joined[joined.columns[~nulls]]
expression_classes = expression.classes.loc[~nulls.reset_index(drop=True)]
expression = joined
mx_diff = 'T' if mx_diff else 'F'
procedure = 'gene_permutation' if single_sample else 'bayes'
cwd = Path(__file__).parent
with NamedTemporaryFile(mode='w', prefix=gsva_tmp_dir) as f_expression, NamedTemporaryFile(prefix=gsva_tmp_dir) as f_result:
expression.to_csv(f_expression)
script = f"""
source("{cwd}/gsva.R")
expression = read.csv('{f_expression.name}', row.names=1)
expression_classes = c{tuple(expression_classes)}
gene_sets = readRDS('{gene_sets_path}')
result = gsva.with_probabilities(
expression, expression_classes, gene_sets, '{procedure}',
method = '{method}', mx.diff={mx_diff}, include_control=F, cores={cores},
limit_to_gene_sets={'c' + str(tuple(limit_to_gene_sets)) if limit_to_gene_sets is not False else 'F'}, progress=F
{', permutations = ' + str(permutations) if procedure == 'permutations' else ''}
)
write.csv(result, '{f_result.name}')
"""
process = Popen(vanilla_R, stdin=PIPE, stdout=PIPE, stderr=PIPE)
r = process._stdin_write(script.encode())
if verbose:
from helpers.streams import handle_streams
from signature_scoring.evaluation import display
handlers = {'out': display, 'err': warn}
handle_streams(process, handlers)
else:
process.wait()
try:
result = read_csv(f_result.name, index_col=0)
except EmptyDataError:
result = DataFrame()
if _cache:
GSVA_CACHE[key] = result
return result