-
Notifications
You must be signed in to change notification settings - Fork 0
/
deepcre_app.py
303 lines (274 loc) · 21.5 KB
/
deepcre_app.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
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
import numpy as np
import pandas as pd
import streamlit as st
from tensorflow.keras.models import load_model
from tensorflow.keras import backend
import os
from utils import prepare_seqs, compute_scores
from io import StringIO
from plotly.subplots import make_subplots
import plotly.graph_objects as go
st.set_page_config(layout='wide')
os.environ['CUDA_VISIBLE_DEVICES'] = '-1'
st.title(':red[DeepCRE Toolkit]')
st.subheader('This tool predicts the probability that a gene will be expressed highly or lowly.')
@st.cache_data
def run_model(data, selected_model):
if data:
input_fasta = StringIO(data.getvalue().decode("utf-8"))
else:
input_fasta = "example.fa"
onehot_seq, seq_ids = prepare_seqs(input_fasta)
model = load_model(saved_models[selected_model])
prediction = model.predict(onehot_seq).ravel()
predicted_class = prediction > 0.5
pred_df = pd.DataFrame({'gene_ids': seq_ids, 'predicted_prob': prediction, 'High expressed': predicted_class})
scores = compute_scores(onehot_data=onehot_seq, keras_model=model)
backend.clear_session()
return scores, pred_df
tab1main, tab2main, tab3main = st.tabs(['Home', 'Predictions', 'Saliency Maps'])
with tab1main:
st.write("""
These models were trained to predict the probability of a gene from a plant species being highly expressed. They expect
as inputs, the flanking regions of genes. We define flanking regions as 1500 nt upstream and 1500 nt downstream of the
your gene of interest. The upstream sequence is obtained by anchoring at the gene start, taking 1000 nt before
and 500 nt after it. This gives you 1500 nt upstream. The downstream sequence is obtained by anchoring on the gene end
taking 500 nt before and 1000 nt after it. This also gives you 1500 nt downstream.\n
Sequences should look like the example sequence GCCGTCCGBREAKGCGGCGCGT. Notice there is a word 'BREAK' in the sequence,
this is not part of the normal DNA sequence, but it tells us were your upstream sequence ends and were the downstream
begins. The sequence in your fasta is\n
\>AT10g0560\n
GCCGTCCGBREAKGCGGCGCGT
If No file is provided the model will run on the default file
""")
col1, _ = st.columns(2)
with col1:
st.image('extraction.png', caption='How to extract flanking sequences')
model_type = st.sidebar.selectbox("""Select a model to use for predictions. The SSR model is trained on Arabidopsis thaliana, while
the Siamese MSR model is trained on 4 species: A. thaliana, S. lycopersicum, S. bicolor and Z. mays.""", options=['SSR', 'Siamese MSR'])
saved_models = {'SSR': 'arabidopsis_model_1_promoter_terminator.h5', 'Siamese MSR': 'siamese_super_msr_model.h5'}
input_fasta = st.sidebar.file_uploader(label="""Upload a fasta file""")
with tab2main:
saliency_scores, meta_df = run_model(input_fasta, model_type)
st.write(meta_df)
with tab3main:
vis_group = st.sidebar.selectbox(label="groupy_by", options=['High vs Low', 'gene_id'])
region = st.sidebar.selectbox(label='Highlight region', options=['UTR', 'Promoter-Terminator'])
rolling_window = st.sidebar.selectbox(label='Choose rolling window', options=[10, 25, 50])
if vis_group == "High vs Low":
tab1_hl, tab2_hl, tab3_hl = st.tabs(['Line', 'Area', 'Histogram'])
# High
mean_high = np.mean(saliency_scores[np.where(meta_df['High expressed'].values == 1)[0]], axis=0)
scores_high = pd.DataFrame({'nucleotide position': np.arange(mean_high.shape[0]),
'A': mean_high[:, 0],
'C': mean_high[:, 1],
'G': mean_high[:, 2],
'T': mean_high[:, 3]})
# Low
mean_low = np.mean(saliency_scores[np.where(meta_df['High expressed'].values == 0)[0]], axis=0)
scores_low = pd.DataFrame({'nucleotide position': np.arange(mean_low.shape[0]),
'A': mean_low[:, 0],
'C': mean_low[:, 1],
'G': mean_low[:, 2],
'T': mean_low[:, 3]})
with tab1_hl:
chart_hl_line = make_subplots(rows=1, cols=2,
subplot_titles=['Shap importance scores: prediction = High expression',
'Shap importance scores: prediction = Low expression'],
shared_yaxes=True)
for group_id, score_df, col in zip(['High', 'Low'], [scores_high, scores_low], [1, 2]):
show_legend = True if group_id == 'High' else False
chart_hl_line.add_trace(go.Scatter(x=score_df['nucleotide position'], y=score_df['A'],
marker=dict(color='green'), name='A',
showlegend=show_legend, legendgroup='A'), row=1, col=col)
chart_hl_line.add_trace(go.Scatter(x=score_df['nucleotide position'], y=score_df['C'],
marker=dict(color='cornflowerblue'), name='C',
showlegend=show_legend, legendgroup='C'), row=1, col=col)
chart_hl_line.add_trace(go.Scatter(x=score_df['nucleotide position'], y=score_df['G'],
marker=dict(color='red'), name='G',
showlegend=show_legend, legendgroup='G'), row=1, col=col)
chart_hl_line.add_trace(go.Scatter(x=score_df['nucleotide position'], y=score_df['T'],
marker=dict(color='darkorange'), name='T',
showlegend=show_legend, legendgroup='T'), row=1, col=col)
chart_hl_line.update_xaxes(title_text="nucleotide position", row=1, col=col)
chart_hl_line.update_yaxes(title_text="saliency score", row=1, col=col)
chart_hl_line.update_xaxes(ticktext=[-1000, -500, 'TSS', 'TTS', 500, 1000],
tickvals=[0, 499, 999, 2019, 2519, 3019])
chart_hl_line.update_yaxes(tickformat=".4f")
if region == 'UTR':
chart_hl_line.add_vrect(x0=999, x1=999 + 500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="5'UTR", annotation_position="top left")
chart_hl_line.add_vrect(x0=1519, x1=1519 + 500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="3'UTR", annotation_position="top left")
else:
chart_hl_line.add_vrect(x0=0, x1=999, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="promoter", annotation_position="top left")
chart_hl_line.add_vrect(x0=1519 + 500, x1=3019, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="terminator", annotation_position="top left")
st.plotly_chart(chart_hl_line, use_container_width=True)
with tab2_hl:
chart_hl_area = make_subplots(rows=1, cols=2,
subplot_titles=[f'Rolling sum, window = {rolling_window}, prediction = High',
f'Rolling sum, window = {rolling_window}, prediction = Low'],
shared_yaxes=True)
for group_id, score_df, colname, col_idx in zip(['High', 'Low'], [scores_high, scores_low], st.columns(2),
[1, 2]):
with colname:
ma = np.nan_to_num(pd.Series(np.mean(score_df[['A', 'C', 'G', 'T']], axis=1)).rolling(rolling_window).sum().values)
chart_hl_area.add_trace(go.Scatter(x=np.arange(len(ma)), y=ma, name=f'{group_id}', fill='tozeroy'),
row=1, col=col_idx)
chart_hl_area.update_xaxes(title_text="nucleotide position", row=1, col=col_idx)
chart_hl_area.update_yaxes(title_text="saliency score", row=1, col=col_idx)
chart_hl_area.update_xaxes(ticktext=[-1000, -500, 'TSS', 'TTS', 500, 1000],
tickvals=[0, 499, 999, 2019, 2519, 3019])
chart_hl_area.update_yaxes(tickformat=".4f")
if region == 'UTR':
chart_hl_area.add_vrect(x0=999, x1=999 + 500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="5'UTR", annotation_position="top left")
chart_hl_area.add_vrect(x0=1519, x1=1519 + 500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="3'UTR", annotation_position="top left")
else:
chart_hl_area.add_vrect(x0=0, x1=999, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="promoter", annotation_position="top left")
chart_hl_area.add_vrect(x0=1519 + 500, x1=3019, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="terminator", annotation_position="top left")
st.plotly_chart(chart_hl_area, use_container_width=True)
with tab3_hl:
chart_hl_hist = make_subplots(rows=1, cols=2, subplot_titles=['Histogram of saliency scores: prediction = High',
'Histogram of saliency scores: prediction = Low'],
shared_yaxes=True)
for group_id, score_df, colname, col_idx in zip(['High', 'Low'], [scores_high, scores_low], st.columns(2),
[1, 2]):
with colname:
chart_hl_hist.add_trace(go.Histogram(x=np.mean(score_df[['A', 'C', 'G', 'T']], axis=1),
name=f'{group_id}'), row=1, col=col_idx)
chart_hl_hist.update_xaxes(title_text="Bins", row=1, col=col_idx)
chart_hl_hist.update_yaxes(title_text="saliency score", row=1, col=col_idx)
chart_hl_hist.update_xaxes(tickformat=".4f")
st.plotly_chart(chart_hl_hist, use_container_width=True)
else:
st.write("""You can group genes together and compare group average saliency scores. This may be useful in
cases were you wanted to compare one gene family to another for example. Or functional comparisons.""")
tab1_gl, tab2_gl = st.tabs(['Line and Area', 'Line and Histogram'])
genes_g1 = st.sidebar.multiselect(label="Group 1", options=meta_df['gene_ids'])
genes_g2 = st.sidebar.multiselect(label="Group 2", options=meta_df['gene_ids'])
with tab1_gl:
if len(genes_g1) != 0 and len(genes_g2) != 0:
idx_genes_g1 = meta_df.index[meta_df['gene_ids'].isin(genes_g1)]
idx_genes_g2 = meta_df.index[meta_df['gene_ids'].isin(genes_g2)]
chart_group = make_subplots(rows=1, cols=2, subplot_titles=['Group 1', 'Group 2'], shared_yaxes=True)
chart2_group = make_subplots(rows=1, cols=2, subplot_titles=['Group 1', 'Group 2'], shared_yaxes=True)
for gene_idxs, group_id, col in zip([idx_genes_g1, idx_genes_g2], ['G1', 'G2'], [1, 2]):
show_legend = True if group_id == 'G1' else False
group_score_mean = np.mean(saliency_scores[gene_idxs], axis=0)
max_score = group_score_mean.max()
scores_df = pd.DataFrame({'nucleotide position': np.arange(group_score_mean.shape[0]),
'A': group_score_mean[:, 0],
'C': group_score_mean[:, 1],
'G': group_score_mean[:, 2],
'T': group_score_mean[:, 3]})
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['A'],
marker=dict(color='green'), name='A',
showlegend=show_legend, legendgroup='A'), row=1, col=col)
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['C'],
marker=dict(color='cornflowerblue'), name='C',
showlegend=show_legend, legendgroup='C'), row=1, col=col)
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['G'],
marker=dict(color='red'), name='G',
showlegend=show_legend, legendgroup='G'), row=1, col=col)
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['T'],
marker=dict(color='darkorange'), name='T',
showlegend=show_legend, legendgroup='T'), row=1, col=col)
chart_group.update_xaxes(title_text="nucleotide position", row=1, col=col)
chart_group.update_yaxes(title_text="saliency score", row=1, col=col)
chart_group.update_xaxes(ticktext=[-1000, -500, 'TSS', 'TTS', 500, 1000],
tickvals=[0, 499, 999, 2019, 2519, 3019])
chart_group.update_yaxes(tickformat=".4f")
if region == 'UTR':
chart_group.add_vrect(x0=999, x1=999+500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="5'UTR", annotation_position="top left")
chart_group.add_vrect(x0=1519, x1=1519+500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="3'UTR", annotation_position="top left")
else:
chart_group.add_vrect(x0=0, x1=999, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="promoter", annotation_position="top left")
chart_group.add_vrect(x0=1519+500, x1=3019, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="terminator", annotation_position="top left")
# For chart 2
ma = np.nan_to_num(pd.Series(np.mean(group_score_mean, axis=1)).rolling(rolling_window).sum().values)
chart2_group.add_trace(go.Scatter(x=np.arange(len(ma)), y=ma, name=f'{group_id}',
fill='tozeroy'), row=1, col=col)
chart2_group.update_xaxes(title_text="nucleotide position", row=1, col=col)
chart2_group.update_yaxes(title_text="saliency score", row=1, col=col)
chart2_group.update_xaxes(ticktext=[-1000, -500, 'TSS', 'TTS', 500, 1000],
tickvals=[0, 499, 999, 2019, 2519, 3019])
chart2_group.update_yaxes(tickformat=".4f")
if region == 'UTR':
chart2_group.add_vrect(x0=999, x1=999+500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="5'UTR", annotation_position="top left")
chart2_group.add_vrect(x0=1519, x1=1519+500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="3'UTR", annotation_position="top left")
else:
chart2_group.add_vrect(x0=0, x1=999, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="promoter", annotation_position="top left")
chart2_group.add_vrect(x0=1519+500, x1=3019, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="terminator", annotation_position="top left")
chart_group.update_layout(title_text="Group saliency scores", showlegend=True)
chart2_group.update_layout(title_text=f"Rolling sum of saliency score, window = {rolling_window}",
showlegend=True)
st.plotly_chart(chart_group, use_container_width=True)
st.plotly_chart(chart2_group, use_container_width=True)
with tab2_gl:
if len(genes_g1) != 0 and len(genes_g2) != 0:
idx_genes_g1 = meta_df.index[meta_df['gene_ids'].isin(genes_g1)]
idx_genes_g2 = meta_df.index[meta_df['gene_ids'].isin(genes_g2)]
chart_group = make_subplots(rows=1, cols=2, subplot_titles=['Group 1', 'Group 2'], shared_yaxes=True)
chart2_group = make_subplots(rows=1, cols=2, subplot_titles=['Group 1', 'Group 2'], shared_yaxes=True)
for gene_idxs, group_id, col, colname in zip([idx_genes_g1, idx_genes_g2], ['G1', 'G2'], [1, 2],
st.columns(2)):
show_legend = True if group_id == 'G1' else False
group_score_mean = np.mean(saliency_scores[gene_idxs], axis=0)
max_score = group_score_mean.max()
scores_df = pd.DataFrame({'nucleotide position': np.arange(group_score_mean.shape[0]),
'A': group_score_mean[:, 0],
'C': group_score_mean[:, 1],
'G': group_score_mean[:, 2],
'T': group_score_mean[:, 3]})
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['A'],
marker=dict(color='green'), name='A',
showlegend=show_legend, legendgroup='A'), row=1, col=col)
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['C'],
marker=dict(color='cornflowerblue'), name='C',
showlegend=show_legend, legendgroup='C'), row=1, col=col)
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['G'],
marker=dict(color='red'), name='G',
showlegend=show_legend, legendgroup='G'), row=1, col=col)
chart_group.add_trace(go.Scatter(x=scores_df['nucleotide position'], y=scores_df['T'],
marker=dict(color='darkorange'), name='T',
showlegend=show_legend, legendgroup='T'), row=1, col=col)
chart_group.update_xaxes(title_text="nucleotide position", row=1, col=col)
chart_group.update_yaxes(title_text="saliency score", row=1, col=col)
chart_group.update_xaxes(ticktext=[-1000, -500, 'TSS', 'TTS', 500, 1000],
tickvals=[0, 499, 999, 2019, 2519, 3019])
chart_group.update_yaxes(tickformat=".4f")
if region == 'UTR':
chart_group.add_vrect(x0=999, x1=999 + 500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="5'UTR", annotation_position="top left")
chart_group.add_vrect(x0=1519, x1=1519 + 500, fillcolor='grey', opacity=0.1, line_width=0,
annotation_text="3'UTR", annotation_position="top left")
else:
chart_group.add_vrect(x0=0, x1=999, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="promoter", annotation_position="top left")
chart_group.add_vrect(x0=1519 + 500, x1=3019, fillcolor='blue', opacity=0.05, line_width=0,
annotation_text="terminator", annotation_position="top left")
# For chart 2
with colname:
chart2_group.add_trace(go.Histogram(x=np.mean(scores_df[['A', 'C', 'G', 'T']], axis=1),
name=f'{group_id}'), row=1, col=col)
chart2_group.update_xaxes(title_text="Bins", row=1, col=col)
chart2_group.update_yaxes(title_text="saliency score", row=1, col=col)
chart2_group.update_xaxes(tickformat=".4f")
chart2_group.update_layout(title_text=f"Histogram saliency scores",
showlegend=True)
st.plotly_chart(chart_group, use_container_width=True)
st.plotly_chart(chart2_group, use_container_width=True)