import logomaker
import numpy as np
import pandas as pd
import re
[docs]def meme_to_homer(meme_path, homer_path, score_power=0.85):
"""
Transfer MEME motif format into Homer motif format.
Based on description here: http://homer.ucsd.edu/homer/motif/creatingCustomMotifs.html
The score_power controls Log odds detection threshold by max_score ** score_power
Parameters
----------
meme_path
Input path in meme format
homer_path
Output path in homer format
score_power
Log odds detection threshold = max_score ** score_power
Returns
-------
"""
if score_power >= 1:
raise ValueError('score_power must < 1')
in_motif = False
records = []
with open(meme_path) as f:
for line in f:
if line.startswith('MOTIF'):
if in_motif:
records.append(cur_char)
cur_char = ''
else:
in_motif = True
cur_char = ''
if not in_motif:
continue
else:
cur_char += line
with open(homer_path, 'w') as f:
for record in records:
lines = record.split('\n')
head_line = lines[0]
matrix_line = []
enter_matrix = False
for line in lines:
if line.startswith('letter-probability matrix'):
enter_matrix = True
continue
ll = line.strip().split(' ')
ll = [i.strip() for i in ll]
if enter_matrix and len(ll) == 4:
matrix_line.append(ll)
else:
continue
matrix = np.array(matrix_line).astype(float)
best_score = np.log(matrix.max(axis=1) / 0.25).sum()
uid = head_line.split(' ')[1]
homer_head_line = f'>\t{uid}\t{best_score ** score_power}\n'
matrix_line = '\n'.join(['\t'.join(line) for line in matrix_line]) + '\n'
f.write(homer_head_line + matrix_line)
return
[docs]def meme_motif_file_to_dict(meme_motif_paths):
if isinstance(meme_motif_paths, str):
meme_motif_paths = [meme_motif_paths]
records = {}
uid_set = set()
for meme_motif_path in meme_motif_paths:
with open(meme_motif_path) as f:
first_line = f.readline()
if not first_line.startswith('MEME version'):
raise ValueError('Input file need to be MEME motif format.')
f.seek(0)
header = True
header_text = ''
motif_tmp_text = ''
for line in f:
if line.startswith('MOTIF'):
# save the previous one first
if motif_tmp_text != '':
records[(uid, name)] = header_text + motif_tmp_text
motif_tmp_text = line
try:
_, uid, name = line.strip('\n').split(' ')
except ValueError:
_, uid = line.strip('\n').split(' ')
name = ''
if uid in uid_set:
raise ValueError(f'Found duplicate motif uid {uid} in file {meme_motif_path}. '
f'Motif uid should be unique across all meme files provided.')
else:
uid_set.add(uid)
header = False
elif header:
header_text += line
else:
motif_tmp_text += line
if motif_tmp_text != '':
records[(uid, name)] = header_text + motif_tmp_text
return records
[docs]def single_meme_txt_to_pfm_df(text, bits_scale=True):
sep = re.compile(r'[01].[0-9]+')
enter = False
pfm_rows = []
for row in text.split('\n'):
if enter:
if row.startswith(' '):
numbers = sep.findall(row)
if len(numbers) == 4:
pfm_rows.append(list(map(float, numbers)))
else:
if row.startswith('letter-probability'):
enter = True
continue
pfm = pd.DataFrame(pfm_rows, columns=['A', 'C', 'G', 'T'])
if bits_scale:
information_content = (pfm * np.log2(pfm / 0.25)).sum(axis=1)
pfm = pfm.multiply(information_content, axis=0)
return pfm
[docs]def meme_to_pfm_dict(meme_motif_paths, bits_scale=True):
records = meme_motif_file_to_dict(meme_motif_paths)
pfm_dict = {}
for (uid, _), text in records.items():
pfm_dict[uid] = single_meme_txt_to_pfm_df(text, bits_scale)
return pfm_dict
[docs]def plot_pfm(pfm, ax=None, logo_kws=None):
_logo_kws = dict(font_name='helvetica',
color_scheme=None,
vpad=.1,
ax=ax,
width=.8)
if logo_kws is not None:
_logo_kws.update(logo_kws)
logo = logomaker.Logo(pfm, **_logo_kws)
logo.ax.set_xlim([-1, len(pfm)])
return logo
[docs]def split_meme_motif_file(meme_motif_paths, output_dir):
"""
Given multi motif meme format file, split into single motif meme format file
Parameters
----------
meme_motif_paths
output_dir
Returns
-------
"""
records = meme_motif_file_to_dict(meme_motif_paths)
motif_file_records = []
for (uid, name), text in records.items():
motif_file_path = output_dir / f'{uid}.meme'
motif_file_records.append([uid, name, motif_file_path])
with open(motif_file_path, 'w') as f:
f.write(text)
return motif_file_records
