/
ratios.py
executable file
·198 lines (166 loc) · 6.81 KB
/
ratios.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
import pandas as pd
import numpy as np
def log_ratios(table_tmp, feature_load, sample_load,
taxa_tmp=None, axis_sort=0, N_show=3, level='lowest'):
""""
Parameters
----------
table_tmp: pandas dataframe - a data table of shape (M,N)
N = Features (i.e. OTUs, metabolites) - cols
M = Samples - index
feature_load: pandas dataframe - a table of shape (P,N)
M = PC - cols
N = Features (i.e. OTUs, metabolites) - index
sample_load: pandas dataframe - a table of shape (P,M)
M = PC - cols
N = Samples - index
taxa_tmp: pandas dataframe - a table of shape (T,M)
Optional; default use table index
T = PC - taxa in feature levels
M = Samples - index
axis_sort: int - level on loadings to use
default is 0
level: str - level of taxon to use
default is lowest can be any taxa level
Returns
-------
log_ratios: pandas dataframe - log ratios of highly loaded features
this is in addition to sample loading PC_axis_sort
metadata catagory and sample IDs
ratios_: list - The ratios to plot
Raises
------
ValueError
Raises an error if axis_sort not in loading axis
`ValueError: axis_sort must be in loading axis (i.e.) str([list])`.
Raises an error if input shape (M,N) where N>M
`ValueError: Data-table contains more samples than features,
most likely your data is transposed`.
References
----------
Examples
--------
>>> from deicode.ratios import log_ratios
>>> from deicode.optspace import OptSpace
>>> from deicode.preprocessing import rclr
>>> import numpy as np
>>> import pandas as pd
rclr preprocessing
data is numpy.ndarray
- a array of counts (samples,features)
(with shape (M,N) where N>M)
>>> from deicode.ratios import log_ratios
>>> from deicode.optspace import OptSpace
>>> from deicode.preprocessing import rclr
>>> import numpy as np
>>> import pandas as pd
>>> data = np.array([[3, 3, 0], [0, 4, 2], [3, 0, 1]])
>>> meta_data = np.array([['C1', 'C2', 'C3']]).T
>>> data = pd.DataFrame(data)
>>> meta_data = pd.DataFrame(meta_data,columns=['Clusters'])
>>> table_rclr = rclr().fit_transform(data)
>>> opt = OptSpace().fit(table_rclr)
>>> sample_weights=pd.DataFrame(opt.sample_weights)
>>> feature_weights=pd.DataFrame(opt.feature_weights)
>>> logdf,ratios = log_ratios(data, meta_data,
sample_weights,
feature_weights)
"""
# make "table"
if taxa_tmp is None:
# level has to be lowest but just ID
level = 'lowest'
cols_ = ['taxonomy_0', 'taxonomy_1', 'taxonomy_2',
'taxonomy_3', 'taxonomy_4', 'taxonomy_5', 'taxonomy_6']
values_ = [str(c) for c in list(table_tmp.columns)]
taxa_tmp = pd.DataFrame(np.array([values_] * 7).T,
index=table_tmp.columns,
columns=cols_)
taxa_tmp_ = True
else:
taxa_tmp = taxa_tmp.astype(str)
taxa_tmp[taxa_tmp == 'nan'] = np.nan
taxa_tmp[taxa_tmp == 'None'] = np.nan
taxa_tmp[taxa_tmp == 'Unassigned'] = '__Unclassified'
taxa_tmp_ = False
if table_tmp.shape[0] > table_tmp.shape[1]:
raise ValueError('Data-table contains more samples than features')
if axis_sort not in feature_load.columns:
raise ValueError(
'The axis given to sort is not in the feature rankings')
# convert unknown taxa to "other"
taxa_tmp = clean_taxa_table(taxa_tmp, taxa_tmp_)
# concat features
feature_taxa = pd.concat([feature_load, taxa_tmp], axis=1).dropna(
subset=[axis_sort]).sort_values(axis_sort, ascending=False)
# level groupby
level_grouping = {level_: feature_taxa.groupby(level_).sum(
).sort_values(axis_sort) for level_ in taxa_tmp.columns}
# get group dicts
top_otus = bin_level_markers(level_grouping, feature_taxa, level, N_show)
# get table of ratios
log_ratios = get_log_ratios(table_tmp, top_otus)
# add that data back to the dicts
log_ratios = pd.concat([sample_load, log_ratios], axis=1)
return log_ratios
def get_taxa(dftmp):
""" add taxa labels to a certain level given """
dftmp.columns = ['kingdom', 'phylum', 'class', 'order',
'family', 'genus', 'species']
return dftmp
def clean_taxa_table(taxa_df, taxa_tmp_):
""" bin taxa to "lowest" level and make unclassified as such"""
mask = np.array([list(taxa_df[l].str.len().values)
for l in taxa_df.columns]).T
mask = pd.DataFrame(mask, taxa_df.index, taxa_df.columns)
taxa_df[mask < 5] = np.nan
lowest_assigned = taxa_df.ffill(axis=1).iloc[:, -1]
if taxa_tmp_:
taxa_df['lowest'] = [
'ID:' + str(y) for x,
y in zip(
lowest_assigned,
lowest_assigned.index)]
else:
taxa_df['lowest'] = [
x.split('__')[1] + '(' + x.split('__')[0] + ')_{ID:' + str(y) + '}'
for x, y in zip(
lowest_assigned,
lowest_assigned.index)]
taxa_df[mask < 5] = '__Unclassified'
return taxa_df
def bin_level_markers(level_gp, taxmatch, level, N_show):
"""otus by taxa level given for N x_n,y_n by feature ranking"""
ratios = {}
for i in range(N_show):
top_ = level_gp[level].iloc[[i]].index[0]
bottom_ = level_gp[level].iloc[[-(i + 1)]].index[0]
top_l = list(taxmatch[taxmatch[level].isin([top_])].index)
bottom_l = list(taxmatch[taxmatch[level].isin([bottom_])].index)
ratios[(top_, bottom_)] = (top_l, bottom_l)
return ratios
def get_log_ratios(tabledf, topd):
""" get log ratios for observed taxa given by bin_level_markers"""
log_ratios = []
for (x_i, y_j), (x_i_features, y_j_features) in topd.items():
if not isinstance(x_i_features, (list,)):
x_i_features = [x_i_features]
if not isinstance(y_j_features, (list,)):
y_j_features = [y_j_features]
p_x_i_y_j = tabledf.loc[:, list(x_i_features) + list(y_j_features)]
x_i_vector = tabledf.loc[:, list(x_i_features)][p_x_i_y_j.T.min() > 0]
y_j_vector = tabledf.loc[:, list(y_j_features)][p_x_i_y_j.T.min() > 0]
tmp_ratio = (np.log(x_i_vector).mean(axis=1) -
np.log(y_j_vector).mean(axis=1))
col_ = [
'log(\dfrac{' +
x_i.replace(
'__',
'') +
'}{' +
y_j.replace(
'__',
'') +
'})']
log_ratios.append(pd.DataFrame(tmp_ratio, columns=col_))
return pd.concat(log_ratios, axis=1)