Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -17,8 +17,6 @@ import pandas as pd
|
|
17 |
import tempfile
|
18 |
import os
|
19 |
from typing import List, Dict, Tuple, Optional, Any
|
20 |
-
import io
|
21 |
-
from io import BytesIO
|
22 |
import seaborn as sns
|
23 |
|
24 |
###############################################################################
|
@@ -55,7 +53,8 @@ def parse_fasta(text):
|
|
55 |
current_sequence = []
|
56 |
for line in text.strip().split('\n'):
|
57 |
line = line.strip()
|
58 |
-
if not line:
|
|
|
59 |
if line.startswith('>'):
|
60 |
if current_header:
|
61 |
sequences.append((current_header, ''.join(current_sequence)))
|
@@ -128,7 +127,8 @@ def compute_positionwise_scores(sequence, shap_values, k=4):
|
|
128 |
|
129 |
def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
130 |
n = len(shap_means)
|
131 |
-
if n == 0:
|
|
|
132 |
if window_size >= n:
|
133 |
return (0, n, float(np.mean(shap_means)))
|
134 |
csum = np.zeros(n + 1, dtype=np.float32)
|
@@ -140,9 +140,11 @@ def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
|
140 |
wsum = csum[start + window_size] - csum[start]
|
141 |
wavg = wsum / window_size
|
142 |
if mode == "max" and wavg > best_avg:
|
143 |
-
best_avg = wavg
|
|
|
144 |
elif mode == "min" and wavg < best_avg:
|
145 |
-
best_avg = wavg
|
|
|
146 |
return (best_start, best_start + window_size, float(best_avg))
|
147 |
|
148 |
###############################################################################
|
@@ -201,9 +203,9 @@ def create_importance_bar_plot(shap_values, kmers, top_k=10):
|
|
201 |
plt.tight_layout()
|
202 |
return fig
|
203 |
|
204 |
-
def plot_shap_histogram(shap_array, title="SHAP Distribution in Region"):
|
205 |
fig, ax = plt.subplots(figsize=(6, 4))
|
206 |
-
ax.hist(shap_array, bins=
|
207 |
ax.axvline(0, color='red', linestyle='--', label='0.0')
|
208 |
ax.set_xlabel("SHAP Value")
|
209 |
ax.set_ylabel("Count")
|
@@ -213,7 +215,8 @@ def plot_shap_histogram(shap_array, title="SHAP Distribution in Region"):
|
|
213 |
return fig
|
214 |
|
215 |
def compute_gc_content(sequence):
|
216 |
-
if not sequence:
|
|
|
217 |
gc_count = sequence.count('G') + sequence.count('C')
|
218 |
return (gc_count / len(sequence)) * 100.0
|
219 |
|
@@ -229,23 +232,24 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text="", window_size=500):
|
|
229 |
with open(file_obj, 'r') as f:
|
230 |
text = f.read()
|
231 |
except Exception as e:
|
232 |
-
return (f"Error reading file: {str(e)}", None, None, None, None)
|
233 |
else:
|
234 |
-
return ("Please provide a FASTA sequence.", None, None, None, None)
|
235 |
|
236 |
sequences = parse_fasta(text)
|
237 |
if not sequences:
|
238 |
-
return ("No valid FASTA sequences found.", None, None, None, None)
|
239 |
header, seq = sequences[0]
|
240 |
|
241 |
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
242 |
try:
|
243 |
-
|
|
|
244 |
model = VirusClassifier(256).to(device)
|
245 |
model.load_state_dict(state_dict)
|
246 |
scaler = joblib.load('scaler.pkl')
|
247 |
except Exception as e:
|
248 |
-
return (f"Error loading model/scaler: {str(e)}", None, None, None, None)
|
249 |
|
250 |
freq_vector = sequence_to_kmer_vector(seq)
|
251 |
scaled_vector = scaler.transform(freq_vector.reshape(1, -1))
|
@@ -280,9 +284,11 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text="", window_size=500):
|
|
280 |
heatmap_fig = plot_linear_heatmap(shap_means, title="Genome-wide SHAP")
|
281 |
heatmap_img = fig_to_image(heatmap_fig)
|
282 |
|
|
|
|
|
283 |
state_dict_out = {"seq": seq, "shap_means": shap_means}
|
284 |
|
285 |
-
return (results_text, bar_img, heatmap_img, state_dict_out, header)
|
286 |
|
287 |
###############################################################################
|
288 |
# 8. SUBREGION ANALYSIS (Gradio Step 2)
|
@@ -290,7 +296,7 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text="", window_size=500):
|
|
290 |
|
291 |
def analyze_subregion(state, header, region_start, region_end):
|
292 |
if not state or "seq" not in state or "shap_means" not in state:
|
293 |
-
return ("No sequence data found. Please run Step 1 first.", None, None)
|
294 |
seq = state["seq"]
|
295 |
shap_means = state["shap_means"]
|
296 |
region_start = int(region_start)
|
@@ -298,7 +304,7 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
298 |
region_start = max(0, min(region_start, len(seq)))
|
299 |
region_end = max(0, min(region_end, len(seq)))
|
300 |
if region_end <= region_start:
|
301 |
-
return ("Invalid region range. End must be > Start.", None, None)
|
302 |
region_seq = seq[region_start:region_end]
|
303 |
region_shap = shap_means[region_start:region_end]
|
304 |
gc_percent = compute_gc_content(region_seq)
|
@@ -324,7 +330,9 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
324 |
heatmap_img = fig_to_image(heatmap_fig)
|
325 |
hist_fig = plot_shap_histogram(region_shap, title="SHAP Distribution in Subregion")
|
326 |
hist_img = fig_to_image(hist_fig)
|
327 |
-
|
|
|
|
|
328 |
|
329 |
###############################################################################
|
330 |
# 9. COMPARISON ANALYSIS FUNCTIONS
|
@@ -476,12 +484,12 @@ def analyze_sequence_comparison(file1, file2, fasta1="", fasta2=""):
|
|
476 |
# Analyze first sequence
|
477 |
res1 = analyze_sequence(file1, top_kmers=10, fasta_text=fasta1, window_size=500)
|
478 |
if isinstance(res1[0], str) and "Error" in res1[0]:
|
479 |
-
return (f"Error in sequence 1: {res1[0]}", None, None)
|
480 |
|
481 |
# Analyze second sequence
|
482 |
res2 = analyze_sequence(file2, top_kmers=10, fasta_text=fasta2, window_size=500)
|
483 |
if isinstance(res2[0], str) and "Error" in res2[0]:
|
484 |
-
return (f"Error in sequence 2: {res2[0]}", None, None)
|
485 |
|
486 |
# Extract SHAP values and sequence info
|
487 |
shap1 = res1[3]["shap_means"]
|
@@ -561,11 +569,12 @@ def analyze_sequence_comparison(file1, file2, fasta1="", fasta2=""):
|
|
561 |
)
|
562 |
hist_img = fig_to_image(hist_fig)
|
563 |
|
564 |
-
|
|
|
565 |
|
566 |
except Exception as e:
|
567 |
error_msg = f"Error during sequence comparison: {str(e)}"
|
568 |
-
return error_msg, None, None
|
569 |
|
570 |
###############################################################################
|
571 |
# 11. GENE FEATURE ANALYSIS
|
@@ -753,13 +762,11 @@ def create_simple_genome_diagram(gene_results: List[Dict[str, Any]], genome_leng
|
|
753 |
# Prepare gene name label
|
754 |
label = str(gene.get('gene_name','?'))
|
755 |
|
756 |
-
#
|
757 |
-
# But if your Pillow version supports font.getsize, you can do:
|
758 |
-
# label_width, label_height = font.getsize(label)
|
759 |
label_mask = font.getmask(label)
|
760 |
label_width, label_height = label_mask.size
|
761 |
|
762 |
-
# Alternate label positions
|
763 |
if idx % 2 == 0:
|
764 |
text_y = line_y - track_height - 15
|
765 |
else:
|
@@ -821,12 +828,10 @@ def create_simple_genome_diagram(gene_results: List[Dict[str, Any]], genome_leng
|
|
821 |
|
822 |
return img
|
823 |
|
824 |
-
|
825 |
-
|
826 |
def analyze_gene_features(sequence_file: str,
|
827 |
-
|
828 |
-
|
829 |
-
|
830 |
"""Analyze SHAP values for each gene feature"""
|
831 |
# First analyze whole sequence
|
832 |
sequence_results = analyze_sequence(sequence_file, top_kmers=10, fasta_text=fasta_text)
|
@@ -980,7 +985,7 @@ with gr.Blocks(css=css) as iface:
|
|
980 |
**Step 3**: Analyze gene features and their contributions.
|
981 |
**Step 4**: Compare sequences and analyze differences.
|
982 |
|
983 |
-
**Color Scale**: Negative SHAP = Blue, Zero = White, Positive = Red.
|
984 |
""")
|
985 |
|
986 |
with gr.Tab("1) Full-Sequence Analysis"):
|
@@ -998,6 +1003,7 @@ with gr.Blocks(css=css) as iface:
|
|
998 |
download_results = gr.File(label="Download Results", visible=False, elem_classes="download-button")
|
999 |
seq_state = gr.State()
|
1000 |
header_state = gr.State()
|
|
|
1001 |
analyze_btn.click(
|
1002 |
analyze_sequence,
|
1003 |
inputs=[file_input, top_k, text_input, win_size],
|
@@ -1019,6 +1025,7 @@ with gr.Blocks(css=css) as iface:
|
|
1019 |
subregion_img = gr.Image(label="Subregion SHAP Heatmap (B-W-R)")
|
1020 |
subregion_hist_img = gr.Image(label="SHAP Distribution (Histogram)")
|
1021 |
download_subregion = gr.File(label="Download Subregion Analysis", visible=False, elem_classes="download-button")
|
|
|
1022 |
region_btn.click(
|
1023 |
analyze_subregion,
|
1024 |
inputs=[seq_state, header_state, region_start, region_end],
|
@@ -1065,8 +1072,8 @@ with gr.Blocks(css=css) as iface:
|
|
1065 |
The sequences will be normalized to the same length for comparison.
|
1066 |
|
1067 |
**Color Scale**:
|
1068 |
-
- Red: Sequence 2
|
1069 |
-
- Blue: Sequence 1
|
1070 |
- White: No substantial difference
|
1071 |
""")
|
1072 |
with gr.Row():
|
@@ -1082,6 +1089,7 @@ with gr.Blocks(css=css) as iface:
|
|
1082 |
diff_heatmap = gr.Image(label="SHAP Difference Heatmap")
|
1083 |
diff_hist = gr.Image(label="Distribution of SHAP Differences")
|
1084 |
download_comparison = gr.File(label="Download Comparison Results", visible=False, elem_classes="download-button")
|
|
|
1085 |
compare_btn.click(
|
1086 |
analyze_sequence_comparison,
|
1087 |
inputs=[file_input1, file_input2, text_input1, text_input2],
|
@@ -1110,4 +1118,4 @@ with gr.Blocks(css=css) as iface:
|
|
1110 |
""")
|
1111 |
|
1112 |
if __name__ == "__main__":
|
1113 |
-
iface.launch()
|
|
|
17 |
import tempfile
|
18 |
import os
|
19 |
from typing import List, Dict, Tuple, Optional, Any
|
|
|
|
|
20 |
import seaborn as sns
|
21 |
|
22 |
###############################################################################
|
|
|
53 |
current_sequence = []
|
54 |
for line in text.strip().split('\n'):
|
55 |
line = line.strip()
|
56 |
+
if not line:
|
57 |
+
continue
|
58 |
if line.startswith('>'):
|
59 |
if current_header:
|
60 |
sequences.append((current_header, ''.join(current_sequence)))
|
|
|
127 |
|
128 |
def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
129 |
n = len(shap_means)
|
130 |
+
if n == 0:
|
131 |
+
return (0, 0, 0.0)
|
132 |
if window_size >= n:
|
133 |
return (0, n, float(np.mean(shap_means)))
|
134 |
csum = np.zeros(n + 1, dtype=np.float32)
|
|
|
140 |
wsum = csum[start + window_size] - csum[start]
|
141 |
wavg = wsum / window_size
|
142 |
if mode == "max" and wavg > best_avg:
|
143 |
+
best_avg = wavg
|
144 |
+
best_start = start
|
145 |
elif mode == "min" and wavg < best_avg:
|
146 |
+
best_avg = wavg
|
147 |
+
best_start = start
|
148 |
return (best_start, best_start + window_size, float(best_avg))
|
149 |
|
150 |
###############################################################################
|
|
|
203 |
plt.tight_layout()
|
204 |
return fig
|
205 |
|
206 |
+
def plot_shap_histogram(shap_array, title="SHAP Distribution in Region", num_bins=30):
|
207 |
fig, ax = plt.subplots(figsize=(6, 4))
|
208 |
+
ax.hist(shap_array, bins=num_bins, color='gray', edgecolor='black')
|
209 |
ax.axvline(0, color='red', linestyle='--', label='0.0')
|
210 |
ax.set_xlabel("SHAP Value")
|
211 |
ax.set_ylabel("Count")
|
|
|
215 |
return fig
|
216 |
|
217 |
def compute_gc_content(sequence):
|
218 |
+
if not sequence:
|
219 |
+
return 0
|
220 |
gc_count = sequence.count('G') + sequence.count('C')
|
221 |
return (gc_count / len(sequence)) * 100.0
|
222 |
|
|
|
232 |
with open(file_obj, 'r') as f:
|
233 |
text = f.read()
|
234 |
except Exception as e:
|
235 |
+
return (f"Error reading file: {str(e)}", None, None, None, None, None)
|
236 |
else:
|
237 |
+
return ("Please provide a FASTA sequence.", None, None, None, None, None)
|
238 |
|
239 |
sequences = parse_fasta(text)
|
240 |
if not sequences:
|
241 |
+
return ("No valid FASTA sequences found.", None, None, None, None, None)
|
242 |
header, seq = sequences[0]
|
243 |
|
244 |
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
|
245 |
try:
|
246 |
+
# IMPORTANT: adjust how you load your model as needed
|
247 |
+
state_dict = torch.load('model.pt', map_location=device)
|
248 |
model = VirusClassifier(256).to(device)
|
249 |
model.load_state_dict(state_dict)
|
250 |
scaler = joblib.load('scaler.pkl')
|
251 |
except Exception as e:
|
252 |
+
return (f"Error loading model/scaler: {str(e)}", None, None, None, None, None)
|
253 |
|
254 |
freq_vector = sequence_to_kmer_vector(seq)
|
255 |
scaled_vector = scaler.transform(freq_vector.reshape(1, -1))
|
|
|
284 |
heatmap_fig = plot_linear_heatmap(shap_means, title="Genome-wide SHAP")
|
285 |
heatmap_img = fig_to_image(heatmap_fig)
|
286 |
|
287 |
+
# You might want to provide a CSV or other data for the 6th return item
|
288 |
+
# Here, we'll simply return None for the file download:
|
289 |
state_dict_out = {"seq": seq, "shap_means": shap_means}
|
290 |
|
291 |
+
return (results_text, bar_img, heatmap_img, state_dict_out, header, None)
|
292 |
|
293 |
###############################################################################
|
294 |
# 8. SUBREGION ANALYSIS (Gradio Step 2)
|
|
|
296 |
|
297 |
def analyze_subregion(state, header, region_start, region_end):
|
298 |
if not state or "seq" not in state or "shap_means" not in state:
|
299 |
+
return ("No sequence data found. Please run Step 1 first.", None, None, None)
|
300 |
seq = state["seq"]
|
301 |
shap_means = state["shap_means"]
|
302 |
region_start = int(region_start)
|
|
|
304 |
region_start = max(0, min(region_start, len(seq)))
|
305 |
region_end = max(0, min(region_end, len(seq)))
|
306 |
if region_end <= region_start:
|
307 |
+
return ("Invalid region range. End must be > Start.", None, None, None)
|
308 |
region_seq = seq[region_start:region_end]
|
309 |
region_shap = shap_means[region_start:region_end]
|
310 |
gc_percent = compute_gc_content(region_seq)
|
|
|
330 |
heatmap_img = fig_to_image(heatmap_fig)
|
331 |
hist_fig = plot_shap_histogram(region_shap, title="SHAP Distribution in Subregion")
|
332 |
hist_img = fig_to_image(hist_fig)
|
333 |
+
|
334 |
+
# For demonstration, returning None for the file download as well
|
335 |
+
return (region_info, heatmap_img, hist_img, None)
|
336 |
|
337 |
###############################################################################
|
338 |
# 9. COMPARISON ANALYSIS FUNCTIONS
|
|
|
484 |
# Analyze first sequence
|
485 |
res1 = analyze_sequence(file1, top_kmers=10, fasta_text=fasta1, window_size=500)
|
486 |
if isinstance(res1[0], str) and "Error" in res1[0]:
|
487 |
+
return (f"Error in sequence 1: {res1[0]}", None, None, None)
|
488 |
|
489 |
# Analyze second sequence
|
490 |
res2 = analyze_sequence(file2, top_kmers=10, fasta_text=fasta2, window_size=500)
|
491 |
if isinstance(res2[0], str) and "Error" in res2[0]:
|
492 |
+
return (f"Error in sequence 2: {res2[0]}", None, None, None)
|
493 |
|
494 |
# Extract SHAP values and sequence info
|
495 |
shap1 = res1[3]["shap_means"]
|
|
|
569 |
)
|
570 |
hist_img = fig_to_image(hist_fig)
|
571 |
|
572 |
+
# Return 4 outputs (text, image, image, and a file or None for the last)
|
573 |
+
return (comparison_text, heatmap_img, hist_img, None)
|
574 |
|
575 |
except Exception as e:
|
576 |
error_msg = f"Error during sequence comparison: {str(e)}"
|
577 |
+
return (error_msg, None, None, None)
|
578 |
|
579 |
###############################################################################
|
580 |
# 11. GENE FEATURE ANALYSIS
|
|
|
762 |
# Prepare gene name label
|
763 |
label = str(gene.get('gene_name','?'))
|
764 |
|
765 |
+
# Fallback for label size
|
|
|
|
|
766 |
label_mask = font.getmask(label)
|
767 |
label_width, label_height = label_mask.size
|
768 |
|
769 |
+
# Alternate label positions
|
770 |
if idx % 2 == 0:
|
771 |
text_y = line_y - track_height - 15
|
772 |
else:
|
|
|
828 |
|
829 |
return img
|
830 |
|
|
|
|
|
831 |
def analyze_gene_features(sequence_file: str,
|
832 |
+
features_file: str,
|
833 |
+
fasta_text: str = "",
|
834 |
+
features_text: str = "") -> Tuple[str, Optional[str], Optional[Image.Image]]:
|
835 |
"""Analyze SHAP values for each gene feature"""
|
836 |
# First analyze whole sequence
|
837 |
sequence_results = analyze_sequence(sequence_file, top_kmers=10, fasta_text=fasta_text)
|
|
|
985 |
**Step 3**: Analyze gene features and their contributions.
|
986 |
**Step 4**: Compare sequences and analyze differences.
|
987 |
|
988 |
+
**Color Scale**: Negative SHAP = Blue, Zero = White, Positive SHAP = Red.
|
989 |
""")
|
990 |
|
991 |
with gr.Tab("1) Full-Sequence Analysis"):
|
|
|
1003 |
download_results = gr.File(label="Download Results", visible=False, elem_classes="download-button")
|
1004 |
seq_state = gr.State()
|
1005 |
header_state = gr.State()
|
1006 |
+
|
1007 |
analyze_btn.click(
|
1008 |
analyze_sequence,
|
1009 |
inputs=[file_input, top_k, text_input, win_size],
|
|
|
1025 |
subregion_img = gr.Image(label="Subregion SHAP Heatmap (B-W-R)")
|
1026 |
subregion_hist_img = gr.Image(label="SHAP Distribution (Histogram)")
|
1027 |
download_subregion = gr.File(label="Download Subregion Analysis", visible=False, elem_classes="download-button")
|
1028 |
+
|
1029 |
region_btn.click(
|
1030 |
analyze_subregion,
|
1031 |
inputs=[seq_state, header_state, region_start, region_end],
|
|
|
1072 |
The sequences will be normalized to the same length for comparison.
|
1073 |
|
1074 |
**Color Scale**:
|
1075 |
+
- Red: Sequence 2 more human-like
|
1076 |
+
- Blue: Sequence 1 more human-like
|
1077 |
- White: No substantial difference
|
1078 |
""")
|
1079 |
with gr.Row():
|
|
|
1089 |
diff_heatmap = gr.Image(label="SHAP Difference Heatmap")
|
1090 |
diff_hist = gr.Image(label="Distribution of SHAP Differences")
|
1091 |
download_comparison = gr.File(label="Download Comparison Results", visible=False, elem_classes="download-button")
|
1092 |
+
|
1093 |
compare_btn.click(
|
1094 |
analyze_sequence_comparison,
|
1095 |
inputs=[file_input1, file_input2, text_input1, text_input2],
|
|
|
1118 |
""")
|
1119 |
|
1120 |
if __name__ == "__main__":
|
1121 |
+
iface.launch()
|