-
Notifications
You must be signed in to change notification settings - Fork 0
/
1.SVDmerge_to_remove_batch_effects.py
280 lines (225 loc) · 9.34 KB
/
1.SVDmerge_to_remove_batch_effects.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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
import numpy as np
import pandas as pd
from sklearn.decomposition import PCA
import matplotlib.pyplot as plt
#get_ipython().magic('matplotlib inline')
import seaborn as sns
from scipy.stats import ks_2samp
import copy
import os
os.chdir("/home/gianmarco/Scrivania/First_Semester/Data_Mining_Laboratory/Project")
human_batches = ["batch1", "batch2", "batch3", "batch4", "batch5", "batch6"]
def normalize_df(df):
"""
input: pandas dataframe, values are log2 gene expression, genes in columns, samples in rows.
output: same, with sum of genes equal 1 for all samples (in linear space).
"""
# exponentiate
x = pow(2.,df)
# rows sum to 1
x = x.div(x.sum(axis=1), axis=0)
return np.log2(x)
dfs = {}
cols = {}
for batch in human_batches:
if batch != "batch6":
dfs[batch] = pd.read_csv("./output/results/preprocessing/"+batch+"_geno.txt", sep="\t", index_col=0)
#cols[batch] = {}
#for col in dfs[batch].columns:
# cols[batch][col] = float
else:
dfs[batch] = pd.read_csv("./output/results/preprocessing/"+batch+"_geno.txt", sep="\t", index_col=0, header="infer").T
#cols[batch] = {}
#for col in dfs[batch].columns:
# cols[batch][col] = float
for batch in dfs:
dfs[batch]=(normalize_df(dfs[batch]))
geno = pd.concat(dfs, axis=1).dropna(axis=0).T
geno.index = geno.index.droplevel(0)
#
dfs = {}
for gse in human_batches:
tmp = pd.read_pickle("./output/results/preprocessing/"+gse+"_pheno.p")
tmp["batch"] = gse
dfs[gse]=(tmp)
pheno = pd.concat(dfs)
pheno.index = pheno.index.droplevel(0)
geno.drop(geno.columns[geno.isnull().sum()!=0],axis=1,inplace=True)
print("batch \t lean \t overw \t obese \t total")
print("---")
for gse in human_batches:
l = ((pheno["batch"]==gse) & (pheno["cbmi"]=="lean")).sum()
over = ((pheno["batch"]==gse) & (pheno["cbmi"]=="overweight")).sum()
o = ((pheno["batch"]==gse) & (pheno["cbmi"]=="obese")).sum()
print("%s \t %d \t %d \t %d \t %d" % (gse,l,over,o,l+over+o))
l = (pheno["cbmi"]=="lean").sum()
over = ((pheno["cbmi"]=="overweight")).sum()
o = ((pheno["cbmi"]=="obese")).sum()
print("---")
print("%s \t %d \t %d \t %d \t %d" % ("ALL\t",l,over,o,l+over+o))
idx = pheno[(pheno["cbmi"]=="overweight")].index
geno.drop(idx,axis=0,inplace=True)
pheno.drop(idx,axis=0,inplace=True)
print("genes:",geno.shape[1])
print("samples (lean):",geno.loc[pheno.cbmi=="lean"].shape[0])
print("samples (obese):",geno.loc[pheno.cbmi=="obese"].shape[0])
ogeno = copy.deepcopy(geno)
def find_level(df,pheno,n=10):
"""
returns 1./pvalue of distro of n first components being from different distros when split by cbmi
args:
df = dataframe with expression values
pheno = dataframe with labels and metadata
n = max numb of components
"""
# make sure we have enough dimensions
n=min(n,df.shape[0])
res = []
pca = PCA()
trans = pca.fit_transform(df)
df_trans = pd.DataFrame(data = trans[:,:n],index = df.index)
for i in range(n):
res.append([i,1/ks_2samp(
df_trans.loc[(pheno.cbmi=="lean") | (pheno.cbmi=="overweight")].iloc[:,i],
df_trans.loc[pheno.cbmi=="obese"].iloc[:,i]
).pvalue])
return np.array(res).T
plt.figure(figsize=(17,3))
for i,batch in enumerate(human_batches):
plt.subplot(1,6,i+1)
plt.title(batch,size=14)
plt.xlabel("eigengene index",size=14)
if i==0: plt.ylabel("$p$-value, KS lean/obese",size=14)
y = np.log10(find_level(geno.loc[pheno.batch==batch],pheno)[1])
plt.plot(range(1,y.shape[0]+1),y)
yticks = plt.yticks()[0][::2]
plt.yticks(yticks,["$10^{-%1.1f}$"%i for i in yticks],fontsize=14)
plt.subplots_adjust(wspace=0.3)
plt.savefig("./output/replication/our/figures/FigureP1.pdf")
pca = PCA()
for batch in human_batches:
effect_strength = find_level(geno.loc[pheno.batch==batch],pheno)[1]
while np.argmax(effect_strength)!=0:
tmp = pca.fit_transform(geno.loc[pheno.batch==batch])
tmp[:,0]=0
print("One principal component was set to zero on batch",batch)
geno.loc[pheno.batch==batch] = pca.inverse_transform(tmp)
effect_strength = find_level(geno.loc[pheno.batch==batch],pheno,n=min(10,geno.loc[pheno.batch==batch].shape[0]))[1]
plt.figure(figsize=(17,3))
for i,batch in enumerate(human_batches):
plt.subplot(1,6,i+1)
plt.title(batch,size=14)
plt.xlabel("eigengene index",size=14)
if i==0: plt.ylabel("$p$-value, KS lean/obese",size=14)
y = np.log10(find_level(geno.loc[pheno.batch==batch],pheno)[1])
plt.plot(range(1,y.shape[0]+1),y)
yticks = plt.yticks()[0][::2]
plt.yticks(yticks,["$10^{-%1.1f}$"%i for i in yticks],fontsize=14)
plt.subplots_adjust(wspace=0.3)
plt.savefig("./output/replication/our/figures/FigureP2.pdf")
plt.title("ALL GSE's",size=14)
plt.plot(range(1,11),np.log10(find_level(geno,pheno)[1]))
plt.xlabel("eigegene index",size=14)
plt.ylabel("$p$-value (KS lean/obese)",size=14)
plt.xticks(fontsize=14)
yticks = plt.yticks()[0][::2]
plt.yticks(yticks,["$10^{-%d}$"%i for i in yticks],fontsize=16)
#plt.show()
eigvect_k = np.argmax(find_level(geno,pheno)[1])
print("Picking eigengene number",eigvect_k+1)
pca = PCA()
trans = pca.fit_transform(geno)
trans[:,:eigvect_k]=0
n_geno = pd.DataFrame(index = geno.index,columns=geno.columns,data=pca.inverse_transform(trans))
K=7
pca_geno = pd.DataFrame(
data = PCA(n_components=K,whiten=True).fit_transform(geno),
index = geno.index).T.corr()
pca_ogeno = pd.DataFrame(
data = PCA(n_components=K,whiten=True).fit_transform(ogeno),
index = ogeno.index).T.corr()
pca_ngeno = pd.DataFrame(
data = PCA(n_components=K,whiten=True).fit_transform(n_geno),
index = ogeno.index).T.corr()
matrix_colors_dict = dict(zip(np.unique(pheno.batch),
np.array([sns.palettes.light_palette(x)[2] for x in sns.color_palette(n_colors=9)])[[3,4,5,6,7, 8]]
))
matrix_colors_dict["lean"] = sns.palettes.light_palette("green")[1]
matrix_colors_dict["obese"] = sns.palettes.light_palette("red")[1]
# sort by batch
order = pheno.sort_values(by=["batch","cbmi"][::1]).index
batchcolors_bybatch = [matrix_colors_dict[pheno.loc[x,"batch"]] for x in order]
cbmicolors_bybatch = [matrix_colors_dict[pheno.loc[x,"cbmi"]] for x in order]
# sort by cbmi
order = pheno.sort_values(by=["batch","cbmi"][::-1]).index
batchcolors_bycbmi = [matrix_colors_dict[pheno.loc[x,"batch"]] for x in order]
cbmicolors_bycbmi = [matrix_colors_dict[pheno.loc[x,"cbmi"]] for x in order]
batch_sizes = np.array([39,20,23,6])
cbmi_size = np.array([39,49])
order = pheno.sort_values(by=["batch","cbmi"][::1]).index
big_ax = sns.clustermap(pca_geno.loc[order,order],
row_cluster=False,col_cluster=False,
row_colors=cbmicolors_bybatch,
col_colors=batchcolors_bybatch,
xticklabels=False,
yticklabels=False,
figsize=(6,6)
)
ax = big_ax.ax_heatmap
big_ax.cax.set_visible(False)
N = n_geno.shape[(0)]
nn=0
for b,n in enumerate(batch_sizes[:-1]):
nn=nn+n
ax.axhline(N-nn,color="black")
ax.axvline(nn,color="black")
ax.text(nn-n/2,N+1,"Batch"+str(b+1),rotation=0,fontsize=13.5,color="black",horizontalalignment="center")
ax.text(N-3,N+1,"B4",rotation=0,fontsize=13.5,color="black",horizontalalignment="center")
#plt.savefig("./output/replication/our/figures/Figure1a.pdf")
plt.show()
order = pheno.sort_values(by=["batch","cbmi"][::1]).index
big_ax = sns.clustermap(pca_ngeno.loc[order,order],
row_cluster=False,col_cluster=False,
row_colors=cbmicolors_bybatch,
col_colors=batchcolors_bybatch,
xticklabels=False,
yticklabels=False,
figsize=(6,6)
)
ax = big_ax.ax_heatmap
big_ax.cax.set_visible(False)
N = n_geno.shape[(0)]
nn=0
for b,n in enumerate(batch_sizes[:-1]):
nn=nn+n
ax.axhline(N-nn,color="black")
ax.axvline(nn,color="black")
ax.text(nn-n/2,N+1,"Batch"+str(b+1),rotation=0,fontsize=13.5,color="black",horizontalalignment="center")
ax.text(N-3,N+1,"B4",rotation=0,fontsize=13.5,color="black",horizontalalignment="center")
#plt.savefig("./output/replication/our/figures/Figure1b.pdf")
plt.show()
order = pheno.sort_values(by=["batch","cbmi"][::-1]).index
big_ax = sns.clustermap(pca_ngeno.loc[order,order],
row_cluster=False,col_cluster=False,
row_colors=cbmicolors_bycbmi,
col_colors=batchcolors_bycbmi,
xticklabels=False,
yticklabels=False,
figsize=(6,6)
)
ax = big_ax.ax_heatmap
big_ax.cax.set_visible(False)
N = n_geno.shape[(0)]
ax.axhline(N-39,color="black")
ax.axvline(39,color="black")
ax.text(-4.5,49/2,"OBESE",rotation=90,fontsize=13.5,color="black",verticalalignment="center")
ax.text(-4.5,N-39/2,"LEAN",rotation=90,fontsize=13.5,color="black",verticalalignment="center")
#lt.savefig("./output/replication/our/figures/Figure1c.pdf")
plt.show()
# ### Save batches 1-4 merged
# Finally, we save the merged batches:
n_geno.to_pickle("./output/results/batch1234_geno.p")
pheno.to_pickle("./output/results/batch1234_pheno.p")
# We have thus removed most of the batch effects, and so batches 1 to 4 have been properly merged.
#
# We can now proceed to identify the relevant genes in [this second notebook](2. Extracting the Obesity Signature.ipynb)