Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -133,7 +133,52 @@ def compute_positionwise_scores(sequence, shap_values, k=4):
|
|
133 |
return shap_means
|
134 |
|
135 |
###############################################################################
|
136 |
-
# 5.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
137 |
###############################################################################
|
138 |
|
139 |
def fig_to_image(fig):
|
@@ -150,7 +195,7 @@ def plot_linear_heatmap(shap_means, title="Per-base SHAP Heatmap", start=None, e
|
|
150 |
Plots a 1D heatmap of per-base SHAP contributions.
|
151 |
Negative = push toward Non-Human, Positive = push toward Human.
|
152 |
Optionally can show only a subrange (start:end).
|
153 |
-
We'll
|
154 |
"""
|
155 |
if start is not None and end is not None:
|
156 |
shap_means = shap_means[start:end]
|
@@ -162,17 +207,17 @@ def plot_linear_heatmap(shap_means, title="Per-base SHAP Heatmap", start=None, e
|
|
162 |
|
163 |
fig, ax = plt.subplots(figsize=(12, 2))
|
164 |
cax = ax.imshow(heatmap_data, aspect='auto', cmap='RdBu_r')
|
165 |
-
|
|
|
|
|
|
|
166 |
cbar.set_label('SHAP Contribution')
|
167 |
|
168 |
ax.set_yticks([])
|
169 |
ax.set_xlabel('Position in Sequence')
|
170 |
ax.set_title(f"{title}{subtitle}")
|
171 |
-
|
172 |
-
|
173 |
-
plt.tight_layout()
|
174 |
-
# Or you can do something like:
|
175 |
-
# plt.subplots_adjust(bottom=0.2)
|
176 |
|
177 |
return fig
|
178 |
|
@@ -219,11 +264,14 @@ def compute_gc_content(sequence):
|
|
219 |
return (gc_count / len(sequence)) * 100.0
|
220 |
|
221 |
###############################################################################
|
222 |
-
#
|
223 |
###############################################################################
|
224 |
|
225 |
-
def analyze_sequence(file_obj, top_kmers=10, fasta_text=""):
|
226 |
-
"""
|
|
|
|
|
|
|
227 |
# Handle input
|
228 |
if fasta_text.strip():
|
229 |
text = fasta_text.strip()
|
@@ -232,14 +280,14 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text=""):
|
|
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)
|
236 |
else:
|
237 |
-
return ("Please provide a FASTA sequence.", None, None, None, None)
|
238 |
|
239 |
# Parse FASTA
|
240 |
sequences = parse_fasta(text)
|
241 |
if not sequences:
|
242 |
-
return ("No valid FASTA sequences found.", None, None, None, None)
|
243 |
|
244 |
header, seq = sequences[0]
|
245 |
|
@@ -250,7 +298,7 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text=""):
|
|
250 |
model.load_state_dict(torch.load('model.pt', map_location=device))
|
251 |
scaler = joblib.load('scaler.pkl')
|
252 |
except Exception as e:
|
253 |
-
return (f"Error loading model: {str(e)}", None, None, None, None)
|
254 |
|
255 |
# Vectorize + scale
|
256 |
freq_vector = sequence_to_kmer_vector(seq)
|
@@ -264,13 +312,26 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text=""):
|
|
264 |
classification = "Human" if prob_human > 0.5 else "Non-human"
|
265 |
confidence = max(prob_human, prob_nonhuman)
|
266 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
267 |
# Build results text
|
268 |
results_text = (
|
269 |
f"Sequence: {header}\n"
|
270 |
f"Length: {len(seq):,} bases\n"
|
271 |
f"Classification: {classification}\n"
|
272 |
f"Confidence: {confidence:.3f}\n"
|
273 |
-
f"(Human Probability: {prob_human:.3f}, Non-human Probability: {prob_nonhuman:.3f})"
|
|
|
|
|
|
|
|
|
|
|
274 |
)
|
275 |
|
276 |
# K-mer importance plot
|
@@ -278,26 +339,27 @@ def analyze_sequence(file_obj, top_kmers=10, fasta_text=""):
|
|
278 |
bar_fig = create_importance_bar_plot(shap_values, kmers, top_kmers)
|
279 |
bar_img = fig_to_image(bar_fig)
|
280 |
|
281 |
-
#
|
282 |
-
shap_means = compute_positionwise_scores(seq, shap_values, k=4)
|
283 |
heatmap_fig = plot_linear_heatmap(shap_means, title="Genome-wide SHAP")
|
284 |
heatmap_img = fig_to_image(heatmap_fig)
|
285 |
|
286 |
# Return:
|
287 |
-
#
|
288 |
-
#
|
289 |
-
#
|
290 |
-
#
|
291 |
-
#
|
|
|
|
|
292 |
state_dict = {
|
293 |
"seq": seq,
|
294 |
"shap_means": shap_means
|
295 |
}
|
296 |
|
297 |
-
return (results_text, bar_img, heatmap_img, state_dict, header)
|
298 |
|
299 |
###############################################################################
|
300 |
-
#
|
301 |
###############################################################################
|
302 |
|
303 |
def analyze_subregion(state, header, region_start, region_end):
|
@@ -333,7 +395,6 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
333 |
negative_fraction = np.mean(region_shap < 0)
|
334 |
|
335 |
# Simple logic-based interpretation
|
336 |
-
# Adjust thresholds as needed
|
337 |
if avg_shap > 0.05:
|
338 |
region_classification = "Likely pushing toward human"
|
339 |
elif avg_shap < -0.05:
|
@@ -368,7 +429,7 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
368 |
|
369 |
|
370 |
###############################################################################
|
371 |
-
#
|
372 |
###############################################################################
|
373 |
|
374 |
css = """
|
@@ -380,7 +441,7 @@ css = """
|
|
380 |
with gr.Blocks(css=css) as iface:
|
381 |
gr.Markdown("""
|
382 |
# Virus Host Classifier (with Interactive Region Viewer)
|
383 |
-
**Step 1**: Predict overall viral sequence origin (human vs non-human)
|
384 |
**Step 2**: Explore subregions to see local SHAP signals, distribution, GC content, etc.
|
385 |
""")
|
386 |
|
@@ -404,26 +465,37 @@ with gr.Blocks(css=css) as iface:
|
|
404 |
step=1,
|
405 |
label="Number of top k-mers to display"
|
406 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
407 |
analyze_btn = gr.Button("Analyze Sequence", variant="primary")
|
408 |
|
409 |
with gr.Column(scale=2):
|
410 |
results_box = gr.Textbox(
|
411 |
-
label="Classification Results", lines=
|
412 |
)
|
413 |
kmer_img = gr.Image(label="Top k-mer SHAP")
|
414 |
genome_img = gr.Image(label="Genome-wide SHAP Heatmap")
|
415 |
|
416 |
# Hidden states that store data for step 2
|
417 |
-
# "seq_state" will hold { seq, shap_means }.
|
418 |
-
# "header_state" is optional meta info
|
419 |
seq_state = gr.State()
|
420 |
header_state = gr.State()
|
421 |
|
422 |
-
# The "analyze_sequence" function returns
|
|
|
|
|
|
|
|
|
|
|
|
|
423 |
analyze_btn.click(
|
424 |
analyze_sequence,
|
425 |
-
inputs=[file_input, top_k, text_input],
|
426 |
-
outputs=[results_box, kmer_img, genome_img, seq_state, header_state]
|
427 |
)
|
428 |
|
429 |
with gr.Tab("2) Subregion Exploration"):
|
@@ -460,6 +532,9 @@ with gr.Blocks(css=css) as iface:
|
|
460 |
- Local SHAP signals (heatmap & histogram)
|
461 |
- GC content, fraction of bases pushing "human" vs "non-human"
|
462 |
- Simple logic-based interpretation based on average SHAP
|
|
|
|
|
|
|
463 |
""")
|
464 |
|
465 |
if __name__ == "__main__":
|
|
|
133 |
return shap_means
|
134 |
|
135 |
###############################################################################
|
136 |
+
# 5. FIND EXTREME SHAP REGIONS
|
137 |
+
###############################################################################
|
138 |
+
|
139 |
+
def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
140 |
+
"""
|
141 |
+
Finds the subregion of length `window_size` that has the maximum
|
142 |
+
(mode="max") or minimum (mode="min") average SHAP.
|
143 |
+
Returns (best_start, best_end, avg_shap).
|
144 |
+
"""
|
145 |
+
n = len(shap_means)
|
146 |
+
if window_size >= n:
|
147 |
+
# If the window is bigger than the entire sequence, return the whole seq
|
148 |
+
avg_val = np.mean(shap_means) if n > 0 else 0.0
|
149 |
+
return (0, n, avg_val)
|
150 |
+
|
151 |
+
# Rolling sum approach
|
152 |
+
csum = np.cumsum(shap_means) # csum[i] = sum of shap_means[0..i-1]
|
153 |
+
# function to compute sum in [start, start+window_size)
|
154 |
+
def window_sum(start):
|
155 |
+
end = start + window_size
|
156 |
+
return csum[end] - csum[start]
|
157 |
+
|
158 |
+
best_start = 0
|
159 |
+
best_avg = None
|
160 |
+
|
161 |
+
# Initialize the best with the first window
|
162 |
+
best_sum = window_sum(0)
|
163 |
+
best_avg = best_sum / window_size
|
164 |
+
best_start = 0
|
165 |
+
|
166 |
+
for start in range(1, n - window_size + 1):
|
167 |
+
wsum = window_sum(start)
|
168 |
+
wavg = wsum / window_size
|
169 |
+
if mode == "max":
|
170 |
+
if wavg > best_avg:
|
171 |
+
best_avg = wavg
|
172 |
+
best_start = start
|
173 |
+
else: # mode == "min"
|
174 |
+
if wavg < best_avg:
|
175 |
+
best_avg = wavg
|
176 |
+
best_start = start
|
177 |
+
|
178 |
+
return (best_start, best_start + window_size, best_avg)
|
179 |
+
|
180 |
+
###############################################################################
|
181 |
+
# 6. PLOTTING / UTILITIES
|
182 |
###############################################################################
|
183 |
|
184 |
def fig_to_image(fig):
|
|
|
195 |
Plots a 1D heatmap of per-base SHAP contributions.
|
196 |
Negative = push toward Non-Human, Positive = push toward Human.
|
197 |
Optionally can show only a subrange (start:end).
|
198 |
+
We'll adjust layout so that the colorbar is below the x-axis and doesn't overlap.
|
199 |
"""
|
200 |
if start is not None and end is not None:
|
201 |
shap_means = shap_means[start:end]
|
|
|
207 |
|
208 |
fig, ax = plt.subplots(figsize=(12, 2))
|
209 |
cax = ax.imshow(heatmap_data, aspect='auto', cmap='RdBu_r')
|
210 |
+
|
211 |
+
# Adjust colorbar with some extra margin
|
212 |
+
# We'll place the colorbar horizontally below
|
213 |
+
cbar = plt.colorbar(cax, orientation='horizontal', pad=0.25)
|
214 |
cbar.set_label('SHAP Contribution')
|
215 |
|
216 |
ax.set_yticks([])
|
217 |
ax.set_xlabel('Position in Sequence')
|
218 |
ax.set_title(f"{title}{subtitle}")
|
219 |
+
# Additional spacing at bottom to avoid overlap
|
220 |
+
plt.subplots_adjust(bottom=0.3)
|
|
|
|
|
|
|
221 |
|
222 |
return fig
|
223 |
|
|
|
264 |
return (gc_count / len(sequence)) * 100.0
|
265 |
|
266 |
###############################################################################
|
267 |
+
# 7. MAIN ANALYSIS STEP (Gradio Step 1)
|
268 |
###############################################################################
|
269 |
|
270 |
+
def analyze_sequence(file_obj, top_kmers=10, fasta_text="", window_size=500):
|
271 |
+
"""
|
272 |
+
Analyzes the entire genome, returning classification, full-genome heatmap,
|
273 |
+
top k-mer bar plot, and identifies subregions with strongest positive/negative push.
|
274 |
+
"""
|
275 |
# Handle input
|
276 |
if fasta_text.strip():
|
277 |
text = fasta_text.strip()
|
|
|
280 |
with open(file_obj, 'r') as f:
|
281 |
text = f.read()
|
282 |
except Exception as e:
|
283 |
+
return (f"Error reading file: {str(e)}", None, None, None, None, None)
|
284 |
else:
|
285 |
+
return ("Please provide a FASTA sequence.", None, None, None, None, None)
|
286 |
|
287 |
# Parse FASTA
|
288 |
sequences = parse_fasta(text)
|
289 |
if not sequences:
|
290 |
+
return ("No valid FASTA sequences found.", None, None, None, None, None)
|
291 |
|
292 |
header, seq = sequences[0]
|
293 |
|
|
|
298 |
model.load_state_dict(torch.load('model.pt', map_location=device))
|
299 |
scaler = joblib.load('scaler.pkl')
|
300 |
except Exception as e:
|
301 |
+
return (f"Error loading model: {str(e)}", None, None, None, None, None)
|
302 |
|
303 |
# Vectorize + scale
|
304 |
freq_vector = sequence_to_kmer_vector(seq)
|
|
|
312 |
classification = "Human" if prob_human > 0.5 else "Non-human"
|
313 |
confidence = max(prob_human, prob_nonhuman)
|
314 |
|
315 |
+
# Per-base SHAP
|
316 |
+
shap_means = compute_positionwise_scores(seq, shap_values, k=4)
|
317 |
+
|
318 |
+
# Find the most "human-pushing" region
|
319 |
+
(max_start, max_end, max_avg) = find_extreme_subregion(shap_means, window_size, mode="max")
|
320 |
+
# Find the most "non-human–pushing" region
|
321 |
+
(min_start, min_end, min_avg) = find_extreme_subregion(shap_means, window_size, mode="min")
|
322 |
+
|
323 |
# Build results text
|
324 |
results_text = (
|
325 |
f"Sequence: {header}\n"
|
326 |
f"Length: {len(seq):,} bases\n"
|
327 |
f"Classification: {classification}\n"
|
328 |
f"Confidence: {confidence:.3f}\n"
|
329 |
+
f"(Human Probability: {prob_human:.3f}, Non-human Probability: {prob_nonhuman:.3f})\n\n"
|
330 |
+
f"---\n"
|
331 |
+
f"**Most Human-Pushing {window_size}-bp Subregion**:\n"
|
332 |
+
f"Start: {max_start}, End: {max_end}, Avg SHAP: {max_avg:.4f}\n\n"
|
333 |
+
f"**Most Non-Human–Pushing {window_size}-bp Subregion**:\n"
|
334 |
+
f"Start: {min_start}, End: {min_end}, Avg SHAP: {min_avg:.4f}"
|
335 |
)
|
336 |
|
337 |
# K-mer importance plot
|
|
|
339 |
bar_fig = create_importance_bar_plot(shap_values, kmers, top_kmers)
|
340 |
bar_img = fig_to_image(bar_fig)
|
341 |
|
342 |
+
# Full-genome SHAP heatmap
|
|
|
343 |
heatmap_fig = plot_linear_heatmap(shap_means, title="Genome-wide SHAP")
|
344 |
heatmap_img = fig_to_image(heatmap_fig)
|
345 |
|
346 |
# Return:
|
347 |
+
# 1) results text
|
348 |
+
# 2) k-mer bar image
|
349 |
+
# 3) full-genome heatmap
|
350 |
+
# 4) "state" with { seq, shap_means, header }, for subregion analysis
|
351 |
+
# 5) we also return "most pushing" subregion info if we want
|
352 |
+
# but for simplicity, we can just keep them in the text.
|
353 |
+
# 6) the sequence header
|
354 |
state_dict = {
|
355 |
"seq": seq,
|
356 |
"shap_means": shap_means
|
357 |
}
|
358 |
|
359 |
+
return (results_text, bar_img, heatmap_img, state_dict, header, None)
|
360 |
|
361 |
###############################################################################
|
362 |
+
# 8. SUBREGION ANALYSIS (Gradio Step 2)
|
363 |
###############################################################################
|
364 |
|
365 |
def analyze_subregion(state, header, region_start, region_end):
|
|
|
395 |
negative_fraction = np.mean(region_shap < 0)
|
396 |
|
397 |
# Simple logic-based interpretation
|
|
|
398 |
if avg_shap > 0.05:
|
399 |
region_classification = "Likely pushing toward human"
|
400 |
elif avg_shap < -0.05:
|
|
|
429 |
|
430 |
|
431 |
###############################################################################
|
432 |
+
# 9. BUILD GRADIO INTERFACE
|
433 |
###############################################################################
|
434 |
|
435 |
css = """
|
|
|
441 |
with gr.Blocks(css=css) as iface:
|
442 |
gr.Markdown("""
|
443 |
# Virus Host Classifier (with Interactive Region Viewer)
|
444 |
+
**Step 1**: Predict overall viral sequence origin (human vs non-human) and identify extreme regions.
|
445 |
**Step 2**: Explore subregions to see local SHAP signals, distribution, GC content, etc.
|
446 |
""")
|
447 |
|
|
|
465 |
step=1,
|
466 |
label="Number of top k-mers to display"
|
467 |
)
|
468 |
+
win_size = gr.Slider(
|
469 |
+
minimum=100,
|
470 |
+
maximum=5000,
|
471 |
+
value=500,
|
472 |
+
step=100,
|
473 |
+
label="Window size for 'most pushing' subregions"
|
474 |
+
)
|
475 |
analyze_btn = gr.Button("Analyze Sequence", variant="primary")
|
476 |
|
477 |
with gr.Column(scale=2):
|
478 |
results_box = gr.Textbox(
|
479 |
+
label="Classification Results", lines=12, interactive=False
|
480 |
)
|
481 |
kmer_img = gr.Image(label="Top k-mer SHAP")
|
482 |
genome_img = gr.Image(label="Genome-wide SHAP Heatmap")
|
483 |
|
484 |
# Hidden states that store data for step 2
|
|
|
|
|
485 |
seq_state = gr.State()
|
486 |
header_state = gr.State()
|
487 |
|
488 |
+
# The "analyze_sequence" function returns 6 values, which we map here:
|
489 |
+
# 1) results_text
|
490 |
+
# 2) bar_img
|
491 |
+
# 3) heatmap_img
|
492 |
+
# 4) state_dict
|
493 |
+
# 5) header
|
494 |
+
# 6) None placeholder
|
495 |
analyze_btn.click(
|
496 |
analyze_sequence,
|
497 |
+
inputs=[file_input, top_k, text_input, win_size],
|
498 |
+
outputs=[results_box, kmer_img, genome_img, seq_state, header_state, None]
|
499 |
)
|
500 |
|
501 |
with gr.Tab("2) Subregion Exploration"):
|
|
|
532 |
- Local SHAP signals (heatmap & histogram)
|
533 |
- GC content, fraction of bases pushing "human" vs "non-human"
|
534 |
- Simple logic-based interpretation based on average SHAP
|
535 |
+
5. **Identification of the most 'human-pushing' subregion** (max average SHAP)
|
536 |
+
and the most 'non-human–pushing' subregion (min average SHAP),
|
537 |
+
each of a chosen window size.
|
538 |
""")
|
539 |
|
540 |
if __name__ == "__main__":
|