import logging
import pandas as pd
import numpy as np
import os
import cmapPy.pandasGEXpress.setup_GCToo_logger as setup_logger
__author__ = "Lev Litichevskiy"
__email__ = "lev@broadinstitute.org"
logger = logging.getLogger(setup_logger.LOGGER_NAME)
# Only writes GCT1.3
VERSION = "1.3"
[docs]def write(gctoo, out_fname, data_null="NaN", metadata_null="-666", filler_null="-666", data_float_format="%.4f"):
"""Write a gctoo object to a gct file.
Args:
gctoo (gctoo object)
out_fname (string): filename for output gct file
data_null (string): how to represent missing values in the data (default = "NaN")
metadata_null (string): how to represent missing values in the metadata (default = "-666")
filler_null (string): what value to fill the top-left filler block with (default = "-666")
data_float_format (string): how many decimal points to keep in representing data
(default = 4 digits; None will keep all digits)
Returns:
None
"""
# Create handle for output file
if not out_fname.endswith(".gct"):
out_fname += ".gct"
f = open(out_fname, "wb")
# Write first two lines
dims = [str(gctoo.data_df.shape[0]), str(gctoo.data_df.shape[1]),
str(gctoo.row_metadata_df.shape[1]), str(gctoo.col_metadata_df.shape[1])]
write_version_and_dims(VERSION, dims, f)
# Write top half of the gct
write_top_half(f, gctoo.row_metadata_df, gctoo.col_metadata_df,
metadata_null, filler_null)
# Write bottom half of the gct
write_bottom_half(f, gctoo.row_metadata_df, gctoo.data_df,
data_null, data_float_format, metadata_null)
f.close()
logger.info("GCT has been written to {}".format(out_fname))
def write_version_and_dims(version, dims, f):
"""Write first two lines of gct file.
Args:
version (string): 1.3 by default
dims (list of strings): length = 4
f (file handle): handle of output file
Returns:
nothing
"""
f.write(("#" + version + "\n"))
f.write((dims[0] + "\t" + dims[1] + "\t" + dims[2] + "\t" + dims[3] + "\n"))
def write_top_half(f, row_metadata_df, col_metadata_df, metadata_null, filler_null):
""" Write the top half of the gct file: top-left filler values, row metadata
headers, and top-right column metadata.
Args:
f (file handle): handle for output file
row_metadata_df (pandas df)
col_metadata_df (pandas df)
metadata_null (string): how to represent missing values in the metadata
filler_null (string): what value to fill the top-left filler block with
Returns:
None
"""
# Initialize the top half of the gct including the third line
size_of_top_half_df = (1 + col_metadata_df.shape[1],
1 + row_metadata_df.shape[1] + col_metadata_df.shape[0])
top_half_df = pd.DataFrame(np.full(size_of_top_half_df, filler_null, dtype=object))
# Assemble the third line of the gct: "id", then rhds, then cids
top_half_df.iloc[0, :] = np.hstack(("id", row_metadata_df.columns.values, col_metadata_df.index.values))
# Insert the chds
top_half_df.iloc[range(1, top_half_df.shape[0]), 0] = col_metadata_df.columns.values
# Insert the column metadata, but first convert to strings and replace NaNs
col_metadata_indices = (range(1, top_half_df.shape[0]),
range(1 + row_metadata_df.shape[1], top_half_df.shape[1]))
top_half_df.iloc[col_metadata_indices[0], col_metadata_indices[1]] = (
col_metadata_df.astype(str).replace("nan", value=metadata_null).T.values)
# Write top_half_df to file
top_half_df.to_csv(f, header=False, index=False, sep="\t")
def write_bottom_half(f, row_metadata_df, data_df, data_null, data_float_format, metadata_null):
""" Write the bottom half of the gct file: row metadata and data.
Args:
f (file handle): handle for output file
row_metadata_df (pandas df)
data_df (pandas df)
data_null (string): how to represent missing values in the data
metadata_null (string): how to represent missing values in the metadata
data_float_format (string): how many decimal points to keep in representing data
Returns:
None
"""
# create the left side of the bottom half of the gct (for the row metadata)
size_of_left_bottom_half_df = (row_metadata_df.shape[0],
1 + row_metadata_df.shape[1])
left_bottom_half_df = pd.DataFrame(np.full(size_of_left_bottom_half_df, metadata_null, dtype=object))
#create the full bottom half by combining with the above with the matrix data
bottom_half_df = pd.concat([left_bottom_half_df, data_df.reset_index(drop=True)], axis=1)
bottom_half_df.columns = range(bottom_half_df.shape[1])
# Insert the rids
bottom_half_df.iloc[:, 0] = row_metadata_df.index.values
# Insert the row metadata, but first convert to strings and replace NaNs
row_metadata_col_indices = range(1, 1 + row_metadata_df.shape[1])
bottom_half_df.iloc[:, row_metadata_col_indices] = (
row_metadata_df.astype(str).replace("nan", value=metadata_null).values)
# Write bottom_half_df to file
bottom_half_df.to_csv(f, header=False, index=False, sep="\t",
na_rep=data_null,
float_format=data_float_format)
def append_dims_and_file_extension(fname, data_df):
"""Append dimensions and file extension to output filename.
N.B. Dimensions are cols x rows.
Args:
fname (string): output filename
data_df (pandas df)
Returns:
out_fname (string): output filename with matrix dims and .gct appended
"""
# If there's no .gct at the end of output file name, add the dims and .gct
if not fname.endswith(".gct"):
out_fname = '{0}_n{1}x{2}.gct'.format(fname, data_df.shape[1], data_df.shape[0])
return out_fname
# Otherwise, only add the dims
else:
basename = os.path.splitext(fname)[0]
out_fname = '{0}_n{1}x{2}.gct'.format(basename, data_df.shape[1], data_df.shape[0])
return out_fname