def calc(df_tpm, genes_for_estimation):
if isinstance(genes_for_estimation, list):
signature_df = df_tpm.ix[genes_for_estimation, [
'reference_sample', 'reference_contamination']]
else:
signature_df = genes_for_estimation
if not isinstance(signature_df, rpy2.robjects.vectors.DataFrame):
signature_df = gg.convert_dataframe_to_r(signature_df, True)
if signature_df is None:
print genes_for_estimation
print type(genes_for_estimation)
raise ValueError("Here it is")
mixture_df = df_tpm[['observed']]
# for some arcane reason it won't work if you run it with one sample
# only...
mixture_df.insert(1, 'duplicate', mixture_df['observed'])
mixture_df = gg.convert_dataframe_to_r(mixture_df, True)
result = rpy2.robjects.r('DeconRNASeq')(
mixture_df,
signature_df,
)
calc_percentage = result[0][1] # the contamination percentage
calc_percentage_std = np.nan
percentage, corrected = correct_with_percentage(
df_tpm,
'observed',
['reference_sample'],
['reference_contamination'],
calc_percentage)
return {
'percentage': calc_percentage,
'percentage_std': calc_percentage_std,
'corrected': corrected,
}
def calc(df_tpm, genes_for_estimation):
rpy2.robjects.r("""call_undo = function(x) {
library("Biobase")
y = as.matrix(x)
colnames(y) = c(1,2)
rownames(y) = rownames(x)
a = unlist(gene_expression_input(y))
markergene <- marker_gene_selection(a,lowper=0.4,highper=0.1,epsilon1=0.01,epsilon2=0.01)
two_source_deconv(y,lowper=0.4,highper=0.1,epsilon1=0.01, epsilon2=0.01,return=0)
}
"""
)
try:
query_df = df_tpm[['observed', 'observed_1']].copy()
# UNDO can't handle zeros in the dataset...
query_df['observed'] += 1
query_df['observed_1'] += 1
x = rpy2.robjects.r('call_undo')(
gg.convert_dataframe_to_r(query_df, True),
)
calc_percentage = np.array(x)[0][0]
percentage, corrected = correct_with_percentage(
df_tpm,
'observed',
['reference_sample'],
['reference_contamination'],
calc_percentage)
except rpy2.rinterface.RRuntimeError as e:
if 'infinite or missing values in' in str(e):
calc_percentage = np.nan
corrected = np.nan
else:
raise
return {
'percentage': calc_percentage,
# 'percentage_std': calc_percentage_std,
'corrected': corrected,
# 'p-value': p_value
}
