Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,5 @@
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
-
import joblib
|
4 |
import numpy as np
|
5 |
from itertools import product
|
6 |
import torch.nn as nn
|
@@ -72,7 +71,7 @@ def sequence_to_kmer_vector(sequence: str, k: int = 4) -> np.ndarray:
|
|
72 |
|
73 |
total_kmers = len(sequence) - k + 1
|
74 |
if total_kmers > 0:
|
75 |
-
vec
|
76 |
|
77 |
return vec
|
78 |
|
@@ -87,12 +86,10 @@ def calculate_shap_values(model, x_tensor):
|
|
87 |
"""
|
88 |
model.eval()
|
89 |
with torch.no_grad():
|
90 |
-
# Baseline
|
91 |
baseline_output = model(x_tensor)
|
92 |
baseline_probs = torch.softmax(baseline_output, dim=1)
|
93 |
baseline_prob = baseline_probs[0, 1].item() # Probability of 'human'
|
94 |
|
95 |
-
# Zeroing each feature to measure impact
|
96 |
shap_values = []
|
97 |
x_zeroed = x_tensor.clone()
|
98 |
for i in range(x_tensor.shape[1]):
|
@@ -100,10 +97,10 @@ def calculate_shap_values(model, x_tensor):
|
|
100 |
x_zeroed[0, i] = 0.0
|
101 |
output = model(x_zeroed)
|
102 |
probs = torch.softmax(output, dim=1)
|
103 |
-
prob = probs[0, 1].item()
|
104 |
impact = baseline_prob - prob
|
105 |
shap_values.append(impact)
|
106 |
-
x_zeroed[0, i] = original_val
|
107 |
return np.array(shap_values), baseline_prob
|
108 |
|
109 |
###############################################################################
|
@@ -111,10 +108,6 @@ def calculate_shap_values(model, x_tensor):
|
|
111 |
###############################################################################
|
112 |
|
113 |
def compute_positionwise_scores(sequence, shap_values, k=4):
|
114 |
-
"""
|
115 |
-
Returns an array of per-base SHAP contributions by averaging
|
116 |
-
the k-mer SHAP values of all k-mers covering that base.
|
117 |
-
"""
|
118 |
kmers = [''.join(p) for p in product("ACGT", repeat=k)]
|
119 |
kmer_dict = {km: i for i, km in enumerate(kmers)}
|
120 |
|
@@ -139,20 +132,13 @@ def compute_positionwise_scores(sequence, shap_values, k=4):
|
|
139 |
###############################################################################
|
140 |
|
141 |
def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
142 |
-
"""
|
143 |
-
Finds the subregion of length `window_size` that has the maximum
|
144 |
-
(mode="max") or minimum (mode="min") average SHAP.
|
145 |
-
Returns (best_start, best_end, best_avg).
|
146 |
-
"""
|
147 |
n = len(shap_means)
|
148 |
if n == 0:
|
149 |
return (0, 0, 0.0)
|
150 |
if window_size >= n:
|
151 |
-
# entire sequence
|
152 |
avg_val = float(np.mean(shap_means))
|
153 |
return (0, n, avg_val)
|
154 |
|
155 |
-
# We'll build csum of length n+1
|
156 |
csum = np.zeros(n + 1, dtype=np.float32)
|
157 |
csum[1:] = np.cumsum(shap_means)
|
158 |
|
@@ -179,7 +165,6 @@ def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
|
179 |
###############################################################################
|
180 |
|
181 |
def fig_to_image(fig):
|
182 |
-
"""Convert a Matplotlib figure to a PIL Image for Gradio."""
|
183 |
buf = io.BytesIO()
|
184 |
fig.savefig(buf, format='png', bbox_inches='tight', dpi=150)
|
185 |
buf.seek(0)
|
@@ -188,27 +173,14 @@ def fig_to_image(fig):
|
|
188 |
return img
|
189 |
|
190 |
def get_zero_centered_cmap():
|
191 |
-
"""
|
192 |
-
Creates a custom diverging colormap that is:
|
193 |
-
- Blue for negative
|
194 |
-
- White for zero
|
195 |
-
- Red for positive
|
196 |
-
"""
|
197 |
colors = [
|
198 |
-
(0.0, 'blue'),
|
199 |
-
(0.5, 'white'),
|
200 |
-
(1.0, 'red')
|
201 |
]
|
202 |
-
|
203 |
-
return cmap
|
204 |
|
205 |
def plot_linear_heatmap(shap_means, title="Per-base SHAP Heatmap", start=None, end=None):
|
206 |
-
"""
|
207 |
-
Plots a 1D heatmap of per-base SHAP contributions with a custom colormap:
|
208 |
-
- Negative = blue
|
209 |
-
- 0 = white
|
210 |
-
- Positive = red
|
211 |
-
"""
|
212 |
if start is not None and end is not None:
|
213 |
local_shap = shap_means[start:end]
|
214 |
subtitle = f" (positions {start}-{end})"
|
@@ -219,73 +191,46 @@ def plot_linear_heatmap(shap_means, title="Per-base SHAP Heatmap", start=None, e
|
|
219 |
if len(local_shap) == 0:
|
220 |
local_shap = np.array([0.0])
|
221 |
|
222 |
-
# Build 2D array for imshow
|
223 |
heatmap_data = local_shap.reshape(1, -1)
|
224 |
-
|
225 |
-
# Force symmetrical range
|
226 |
min_val = np.min(local_shap)
|
227 |
max_val = np.max(local_shap)
|
228 |
extent = max(abs(min_val), abs(max_val))
|
|
|
229 |
|
230 |
-
|
231 |
-
custom_cmap = get_zero_centered_cmap()
|
232 |
-
|
233 |
-
# Create figure with adjusted height ratio
|
234 |
-
fig, ax = plt.subplots(figsize=(12, 1.8)) # Reduced height
|
235 |
-
|
236 |
-
# Plot heatmap
|
237 |
cax = ax.imshow(
|
238 |
heatmap_data,
|
239 |
aspect='auto',
|
240 |
-
cmap=
|
241 |
vmin=-extent,
|
242 |
-
vmax
|
243 |
)
|
244 |
-
|
245 |
-
# Configure colorbar with more subtle positioning
|
246 |
cbar = plt.colorbar(
|
247 |
cax,
|
248 |
orientation='horizontal',
|
249 |
-
pad=0.25,
|
250 |
-
aspect=40,
|
251 |
-
shrink=0.8
|
252 |
-
)
|
253 |
-
|
254 |
-
# Style the colorbar
|
255 |
-
cbar.ax.tick_params(labelsize=8) # Smaller tick labels
|
256 |
-
cbar.set_label(
|
257 |
-
'SHAP Contribution',
|
258 |
-
fontsize=9,
|
259 |
-
labelpad=5
|
260 |
)
|
|
|
|
|
261 |
|
262 |
-
# Configure main plot
|
263 |
ax.set_yticks([])
|
264 |
ax.set_xlabel('Position in Sequence', fontsize=10)
|
265 |
ax.set_title(f"{title}{subtitle}", pad=10)
|
266 |
-
|
267 |
-
# Fine-tune layout
|
268 |
-
plt.subplots_adjust(
|
269 |
-
bottom=0.25, # Reduced bottom margin
|
270 |
-
left=0.05, # Tighter left margin
|
271 |
-
right=0.95 # Tighter right margin
|
272 |
-
)
|
273 |
|
274 |
return fig
|
275 |
|
276 |
def create_importance_bar_plot(shap_values, kmers, top_k=10):
|
277 |
-
"""Create a bar plot of the most important k-mers."""
|
278 |
plt.rcParams.update({'font.size': 10})
|
279 |
fig = plt.figure(figsize=(10, 5))
|
280 |
|
281 |
-
# Sort by absolute importance
|
282 |
indices = np.argsort(np.abs(shap_values))[-top_k:]
|
283 |
values = shap_values[indices]
|
284 |
features = [kmers[i] for i in indices]
|
285 |
|
286 |
-
# negative -> blue, positive -> red
|
287 |
colors = ['#99ccff' if v < 0 else '#ff9999' for v in values]
|
288 |
-
|
289 |
plt.barh(range(len(values)), values, color=colors)
|
290 |
plt.yticks(range(len(values)), features)
|
291 |
plt.xlabel('SHAP Value (impact on model output)')
|
@@ -295,9 +240,6 @@ def create_importance_bar_plot(shap_values, kmers, top_k=10):
|
|
295 |
return fig
|
296 |
|
297 |
def plot_shap_histogram(shap_array, title="SHAP Distribution in Region"):
|
298 |
-
"""
|
299 |
-
Simple histogram of SHAP values in the subregion.
|
300 |
-
"""
|
301 |
fig, ax = plt.subplots(figsize=(6, 4))
|
302 |
ax.hist(shap_array, bins=30, color='gray', edgecolor='black')
|
303 |
ax.axvline(0, color='red', linestyle='--', label='0.0')
|
@@ -309,7 +251,6 @@ def plot_shap_histogram(shap_array, title="SHAP Distribution in Region"):
|
|
309 |
return fig
|
310 |
|
311 |
def compute_gc_content(sequence):
|
312 |
-
"""Compute %GC in the sequence (A, C, G, T)."""
|
313 |
if not sequence:
|
314 |
return 0
|
315 |
gc_count = sequence.count('G') + sequence.count('C')
|
@@ -319,78 +260,72 @@ def compute_gc_content(sequence):
|
|
319 |
# 7. SEQUENCE ANALYSIS FUNCTIONS
|
320 |
###############################################################################
|
321 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
322 |
def analyze_sequence(file_path, top_k=10, fasta_text="", window_size=500):
|
323 |
"""
|
324 |
Analyze a virus sequence from a FASTA file or text input.
|
325 |
Returns (results_text, kmer_plot, heatmap_plot, state_dict, header)
|
326 |
"""
|
327 |
try:
|
328 |
-
# Load model and k-mer info
|
329 |
-
model = VirusClassifier(256) # 4^4 = 256 k-mers for k=4
|
330 |
-
model.load_state_dict(torch.load("model.pt"))
|
331 |
-
model.eval()
|
332 |
-
kmers = [''.join(p) for p in product("ACGT", repeat=4)]
|
333 |
-
|
334 |
-
# Process input (file takes precedence over text)
|
335 |
if file_path:
|
336 |
with open(file_path, 'r') as f:
|
337 |
fasta_text = f.read()
|
338 |
|
339 |
if not fasta_text.strip():
|
340 |
return ("Error: No sequence provided", None, None, {}, "")
|
341 |
-
|
342 |
-
# Parse FASTA
|
343 |
sequences = parse_fasta(fasta_text)
|
344 |
if not sequences:
|
345 |
return ("Error: No valid FASTA sequences found", None, None, {}, "")
|
346 |
-
|
347 |
-
header, sequence = sequences[0] # Take first sequence
|
348 |
|
349 |
-
|
350 |
-
|
351 |
-
|
|
|
352 |
|
353 |
-
# Get model prediction
|
354 |
with torch.no_grad():
|
355 |
output = model(x_tensor)
|
356 |
probs = torch.softmax(output, dim=1)
|
357 |
-
# Using index 1 for probability of human
|
358 |
pred_human = probs[0, 1].item()
|
359 |
-
|
360 |
-
# Calculate SHAP values
|
361 |
-
shap_values, prob = calculate_shap_values(model, x_tensor)
|
362 |
|
363 |
-
|
364 |
-
|
|
|
|
|
|
|
|
|
365 |
start_max, end_max, avg_max = find_extreme_subregion(shap_means, window_size, mode="max")
|
366 |
start_min, end_min, avg_min = find_extreme_subregion(shap_means, window_size, mode="min")
|
367 |
|
368 |
-
# Format results text
|
369 |
-
classification = "Human" if pred_human > 0.5 else "Non-human"
|
370 |
results = (
|
371 |
f"Classification: {classification} "
|
372 |
f"(probability of human = {pred_human:.3f})\n\n"
|
373 |
f"Sequence length: {len(sequence):,} bases\n"
|
374 |
f"Overall GC content: {compute_gc_content(sequence):.1f}%\n\n"
|
375 |
-
f"Most human-like {window_size}bp region:\n"
|
376 |
f"Position {start_max:,} to {end_max:,}\n"
|
377 |
f"Average SHAP: {avg_max:.4f}\n"
|
378 |
f"GC content: {compute_gc_content(sequence[start_max:end_max]):.1f}%\n\n"
|
379 |
-
f"Least human-like {window_size}bp region:\n"
|
380 |
f"Position {start_min:,} to {end_min:,}\n"
|
381 |
f"Average SHAP: {avg_min:.4f}\n"
|
382 |
f"GC content: {compute_gc_content(sequence[start_min:end_min]):.1f}%"
|
383 |
)
|
384 |
|
385 |
-
|
386 |
-
kmer_fig = create_importance_bar_plot(shap_values, kmers, top_k)
|
387 |
kmer_img = fig_to_image(kmer_fig)
|
388 |
|
389 |
-
# Create genome-wide heatmap
|
390 |
heatmap_fig = plot_linear_heatmap(shap_means)
|
391 |
heatmap_img = fig_to_image(heatmap_fig)
|
392 |
|
393 |
-
# Store data for subregion analysis
|
394 |
state = {
|
395 |
"seq": sequence,
|
396 |
"shap_means": shap_means
|
@@ -399,21 +334,19 @@ def analyze_sequence(file_path, top_k=10, fasta_text="", window_size=500):
|
|
399 |
return results, kmer_img, heatmap_img, state, header
|
400 |
|
401 |
except Exception as e:
|
402 |
-
|
403 |
-
|
|
|
|
|
|
|
404 |
|
405 |
def analyze_subregion(state, header, region_start, region_end):
|
406 |
-
"""
|
407 |
-
Takes stored data from step 1 and a user-chosen region.
|
408 |
-
Returns a subregion heatmap, histogram, and some stats (GC, average SHAP).
|
409 |
-
"""
|
410 |
if not state or "seq" not in state or "shap_means" not in state:
|
411 |
return ("No sequence data found. Please run Step 1 first.", None, None)
|
412 |
|
413 |
seq = state["seq"]
|
414 |
shap_means = state["shap_means"]
|
415 |
|
416 |
-
# Validate bounds
|
417 |
region_start = int(region_start)
|
418 |
region_end = int(region_end)
|
419 |
|
@@ -422,19 +355,15 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
422 |
if region_end <= region_start:
|
423 |
return ("Invalid region range. End must be > Start.", None, None)
|
424 |
|
425 |
-
# Subsequence
|
426 |
region_seq = seq[region_start:region_end]
|
427 |
region_shap = shap_means[region_start:region_end]
|
428 |
|
429 |
-
# Some stats
|
430 |
gc_percent = compute_gc_content(region_seq)
|
431 |
avg_shap = float(np.mean(region_shap))
|
432 |
|
433 |
-
# Fraction pushing toward human vs. non-human
|
434 |
positive_fraction = np.mean(region_shap > 0)
|
435 |
negative_fraction = np.mean(region_shap < 0)
|
436 |
|
437 |
-
# Simple logic-based interpretation
|
438 |
if avg_shap > 0.05:
|
439 |
region_classification = "Likely pushing toward human"
|
440 |
elif avg_shap < -0.05:
|
@@ -452,7 +381,6 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
452 |
f"Subregion interpretation: {region_classification}\n"
|
453 |
)
|
454 |
|
455 |
-
# Plot region as small heatmap
|
456 |
heatmap_fig = plot_linear_heatmap(
|
457 |
shap_means,
|
458 |
title="Subregion SHAP",
|
@@ -461,72 +389,45 @@ def analyze_subregion(state, header, region_start, region_end):
|
|
461 |
)
|
462 |
heatmap_img = fig_to_image(heatmap_fig)
|
463 |
|
464 |
-
# Plot histogram of SHAP in region
|
465 |
hist_fig = plot_shap_histogram(region_shap, title="SHAP Distribution in Subregion")
|
466 |
hist_img = fig_to_image(hist_fig)
|
467 |
|
468 |
return (region_info, heatmap_img, hist_img)
|
469 |
|
470 |
###############################################################################
|
471 |
-
#
|
472 |
###############################################################################
|
473 |
|
474 |
def normalize_shap_lengths(shap1, shap2, num_points=1000):
|
475 |
-
"""
|
476 |
-
Normalize two SHAP arrays to the same length using interpolation.
|
477 |
-
Returns (normalized_shap1, normalized_shap2)
|
478 |
-
"""
|
479 |
-
# Create x coordinates for both sequences
|
480 |
x1 = np.linspace(0, 1, len(shap1))
|
481 |
x2 = np.linspace(0, 1, len(shap2))
|
482 |
|
483 |
-
# Create interpolation functions
|
484 |
f1 = interp1d(x1, shap1, kind='linear')
|
485 |
f2 = interp1d(x2, shap2, kind='linear')
|
486 |
|
487 |
-
# Create new x coordinates for interpolation
|
488 |
x_new = np.linspace(0, 1, num_points)
|
489 |
|
490 |
-
# Interpolate both sequences to new length
|
491 |
shap1_norm = f1(x_new)
|
492 |
shap2_norm = f2(x_new)
|
493 |
|
494 |
return shap1_norm, shap2_norm
|
495 |
|
496 |
def compute_shap_difference(shap1_norm, shap2_norm):
|
497 |
-
"""
|
498 |
-
Compute the difference between two normalized SHAP arrays.
|
499 |
-
Positive values indicate seq2 is more "human-like" than seq1.
|
500 |
-
"""
|
501 |
return shap2_norm - shap1_norm
|
502 |
|
503 |
def plot_comparative_heatmap(shap_diff, title="SHAP Difference Heatmap"):
|
504 |
-
"""
|
505 |
-
Plot the difference between two sequences' SHAP values.
|
506 |
-
Red indicates seq2 is more human-like, blue indicates seq1 is more human-like.
|
507 |
-
"""
|
508 |
-
# Build 2D array for imshow
|
509 |
heatmap_data = shap_diff.reshape(1, -1)
|
510 |
-
|
511 |
-
# Force symmetrical range
|
512 |
extent = max(abs(np.min(shap_diff)), abs(np.max(shap_diff)))
|
|
|
513 |
|
514 |
-
# Create figure with adjusted height ratio
|
515 |
fig, ax = plt.subplots(figsize=(12, 1.8))
|
516 |
-
|
517 |
-
# Create custom colormap
|
518 |
-
custom_cmap = get_zero_centered_cmap()
|
519 |
-
|
520 |
-
# Plot heatmap
|
521 |
cax = ax.imshow(
|
522 |
heatmap_data,
|
523 |
aspect='auto',
|
524 |
-
cmap=
|
525 |
vmin=-extent,
|
526 |
-
vmax
|
527 |
)
|
528 |
-
|
529 |
-
# Configure colorbar
|
530 |
cbar = plt.colorbar(
|
531 |
cax,
|
532 |
orientation='horizontal',
|
@@ -534,74 +435,47 @@ def plot_comparative_heatmap(shap_diff, title="SHAP Difference Heatmap"):
|
|
534 |
aspect=40,
|
535 |
shrink=0.8
|
536 |
)
|
537 |
-
|
538 |
-
# Style the colorbar
|
539 |
cbar.ax.tick_params(labelsize=8)
|
540 |
-
cbar.set_label(
|
541 |
-
'SHAP Difference (Seq2 - Seq1)',
|
542 |
-
fontsize=9,
|
543 |
-
labelpad=5
|
544 |
-
)
|
545 |
|
546 |
-
# Configure main plot
|
547 |
ax.set_yticks([])
|
548 |
ax.set_xlabel('Normalized Position (0-100%)', fontsize=10)
|
549 |
ax.set_title(title, pad=10)
|
550 |
-
|
551 |
-
plt.subplots_adjust(
|
552 |
-
bottom=0.25,
|
553 |
-
left=0.05,
|
554 |
-
right=0.95
|
555 |
-
)
|
556 |
|
557 |
return fig
|
558 |
|
559 |
def analyze_sequence_comparison(file1, file2, fasta1="", fasta2=""):
|
560 |
-
|
561 |
-
Compare two sequences by analyzing their SHAP differences.
|
562 |
-
Returns comparison text and visualizations.
|
563 |
-
"""
|
564 |
-
# Process first sequence
|
565 |
-
results1 = analyze_sequence(file1, fasta_text=fasta1)
|
566 |
if isinstance(results1[0], str) and "Error" in results1[0]:
|
567 |
return (f"Error in sequence 1: {results1[0]}", None, None)
|
568 |
|
569 |
-
|
570 |
-
results2 = analyze_sequence(file2, fasta_text=fasta2)
|
571 |
if isinstance(results2[0], str) and "Error" in results2[0]:
|
572 |
return (f"Error in sequence 2: {results2[0]}", None, None)
|
573 |
|
574 |
-
# Get SHAP means from state dictionaries
|
575 |
shap1 = results1[3]["shap_means"]
|
576 |
shap2 = results2[3]["shap_means"]
|
577 |
|
578 |
-
# Normalize lengths
|
579 |
shap1_norm, shap2_norm = normalize_shap_lengths(shap1, shap2)
|
580 |
-
|
581 |
-
# Compute difference (positive = seq2 more human-like)
|
582 |
shap_diff = compute_shap_difference(shap1_norm, shap2_norm)
|
583 |
|
584 |
-
# Calculate statistics
|
585 |
avg_diff = np.mean(shap_diff)
|
586 |
std_diff = np.std(shap_diff)
|
587 |
max_diff = np.max(shap_diff)
|
588 |
min_diff = np.min(shap_diff)
|
589 |
|
590 |
-
|
591 |
-
threshold = 0.05 # Arbitrary threshold for "substantial" difference
|
592 |
substantial_diffs = np.abs(shap_diff) > threshold
|
593 |
frac_different = np.mean(substantial_diffs)
|
594 |
|
595 |
-
# Extract classifications safely
|
596 |
classification1 = results1[0].split('Classification: ')[1].split('\n')[0].strip()
|
597 |
classification2 = results2[0].split('Classification: ')[1].split('\n')[0].strip()
|
598 |
|
599 |
-
# Format numbers
|
600 |
len1_formatted = "{:,}".format(len(shap1))
|
601 |
len2_formatted = "{:,}".format(len(shap2))
|
602 |
frac_formatted = "{:.2%}".format(frac_different)
|
603 |
|
604 |
-
# Build comparison text
|
605 |
comparison_text = (
|
606 |
"Sequence Comparison Results:\n"
|
607 |
f"Sequence 1: {results1[4]}\n"
|
@@ -621,21 +495,16 @@ def analyze_sequence_comparison(file1, file2, fasta1="", fasta2=""):
|
|
621 |
"Negative values (blue) indicate regions where Sequence 1 is more 'human-like'"
|
622 |
)
|
623 |
|
624 |
-
# Create comparison heatmap
|
625 |
heatmap_fig = plot_comparative_heatmap(shap_diff)
|
626 |
heatmap_img = fig_to_image(heatmap_fig)
|
627 |
|
628 |
-
|
629 |
-
hist_fig = plot_shap_histogram(
|
630 |
-
shap_diff,
|
631 |
-
title="Distribution of SHAP Differences"
|
632 |
-
)
|
633 |
hist_img = fig_to_image(hist_fig)
|
634 |
|
635 |
return comparison_text, heatmap_img, hist_img
|
636 |
|
637 |
###############################################################################
|
638 |
-
#
|
639 |
###############################################################################
|
640 |
|
641 |
css = """
|
@@ -666,14 +535,14 @@ with gr.Blocks(css=css) as iface:
|
|
666 |
placeholder=">sequence_name\nACGTACGT...",
|
667 |
lines=5
|
668 |
)
|
669 |
-
|
670 |
minimum=5,
|
671 |
maximum=30,
|
672 |
value=10,
|
673 |
step=1,
|
674 |
label="Number of top k-mers to display"
|
675 |
)
|
676 |
-
|
677 |
minimum=100,
|
678 |
maximum=5000,
|
679 |
value=500,
|
@@ -694,7 +563,7 @@ with gr.Blocks(css=css) as iface:
|
|
694 |
|
695 |
analyze_btn.click(
|
696 |
analyze_sequence,
|
697 |
-
inputs=[file_input,
|
698 |
outputs=[results_box, kmer_img, genome_img, seq_state, header_state]
|
699 |
)
|
700 |
|
@@ -797,22 +666,16 @@ with gr.Blocks(css=css) as iface:
|
|
797 |
- Statistical summary of differences
|
798 |
""")
|
799 |
|
800 |
-
###############################################################################
|
801 |
-
# 10. MAIN EXECUTION
|
802 |
-
###############################################################################
|
803 |
-
|
804 |
if __name__ == "__main__":
|
805 |
-
# Set up any global configurations if needed
|
806 |
plt.style.use('default')
|
807 |
plt.rcParams['figure.figsize'] = [10, 6]
|
808 |
plt.rcParams['figure.dpi'] = 100
|
809 |
plt.rcParams['font.size'] = 10
|
810 |
|
811 |
-
# Launch the interface
|
812 |
iface.launch(
|
813 |
-
share=False,
|
814 |
-
server_name="0.0.0.0",
|
815 |
-
server_port=7860,
|
816 |
-
show_api=False,
|
817 |
-
debug=False
|
818 |
)
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
|
|
3 |
import numpy as np
|
4 |
from itertools import product
|
5 |
import torch.nn as nn
|
|
|
71 |
|
72 |
total_kmers = len(sequence) - k + 1
|
73 |
if total_kmers > 0:
|
74 |
+
vec /= total_kmers
|
75 |
|
76 |
return vec
|
77 |
|
|
|
86 |
"""
|
87 |
model.eval()
|
88 |
with torch.no_grad():
|
|
|
89 |
baseline_output = model(x_tensor)
|
90 |
baseline_probs = torch.softmax(baseline_output, dim=1)
|
91 |
baseline_prob = baseline_probs[0, 1].item() # Probability of 'human'
|
92 |
|
|
|
93 |
shap_values = []
|
94 |
x_zeroed = x_tensor.clone()
|
95 |
for i in range(x_tensor.shape[1]):
|
|
|
97 |
x_zeroed[0, i] = 0.0
|
98 |
output = model(x_zeroed)
|
99 |
probs = torch.softmax(output, dim=1)
|
100 |
+
prob = probs[0, 1].item()
|
101 |
impact = baseline_prob - prob
|
102 |
shap_values.append(impact)
|
103 |
+
x_zeroed[0, i] = original_val
|
104 |
return np.array(shap_values), baseline_prob
|
105 |
|
106 |
###############################################################################
|
|
|
108 |
###############################################################################
|
109 |
|
110 |
def compute_positionwise_scores(sequence, shap_values, k=4):
|
|
|
|
|
|
|
|
|
111 |
kmers = [''.join(p) for p in product("ACGT", repeat=k)]
|
112 |
kmer_dict = {km: i for i, km in enumerate(kmers)}
|
113 |
|
|
|
132 |
###############################################################################
|
133 |
|
134 |
def find_extreme_subregion(shap_means, window_size=500, mode="max"):
|
|
|
|
|
|
|
|
|
|
|
135 |
n = len(shap_means)
|
136 |
if n == 0:
|
137 |
return (0, 0, 0.0)
|
138 |
if window_size >= n:
|
|
|
139 |
avg_val = float(np.mean(shap_means))
|
140 |
return (0, n, avg_val)
|
141 |
|
|
|
142 |
csum = np.zeros(n + 1, dtype=np.float32)
|
143 |
csum[1:] = np.cumsum(shap_means)
|
144 |
|
|
|
165 |
###############################################################################
|
166 |
|
167 |
def fig_to_image(fig):
|
|
|
168 |
buf = io.BytesIO()
|
169 |
fig.savefig(buf, format='png', bbox_inches='tight', dpi=150)
|
170 |
buf.seek(0)
|
|
|
173 |
return img
|
174 |
|
175 |
def get_zero_centered_cmap():
|
|
|
|
|
|
|
|
|
|
|
|
|
176 |
colors = [
|
177 |
+
(0.0, 'blue'),
|
178 |
+
(0.5, 'white'),
|
179 |
+
(1.0, 'red')
|
180 |
]
|
181 |
+
return mcolors.LinearSegmentedColormap.from_list("blue_white_red", colors)
|
|
|
182 |
|
183 |
def plot_linear_heatmap(shap_means, title="Per-base SHAP Heatmap", start=None, end=None):
|
|
|
|
|
|
|
|
|
|
|
|
|
184 |
if start is not None and end is not None:
|
185 |
local_shap = shap_means[start:end]
|
186 |
subtitle = f" (positions {start}-{end})"
|
|
|
191 |
if len(local_shap) == 0:
|
192 |
local_shap = np.array([0.0])
|
193 |
|
|
|
194 |
heatmap_data = local_shap.reshape(1, -1)
|
|
|
|
|
195 |
min_val = np.min(local_shap)
|
196 |
max_val = np.max(local_shap)
|
197 |
extent = max(abs(min_val), abs(max_val))
|
198 |
+
cmap = get_zero_centered_cmap()
|
199 |
|
200 |
+
fig, ax = plt.subplots(figsize=(12, 1.8))
|
|
|
|
|
|
|
|
|
|
|
|
|
201 |
cax = ax.imshow(
|
202 |
heatmap_data,
|
203 |
aspect='auto',
|
204 |
+
cmap=cmap,
|
205 |
vmin=-extent,
|
206 |
+
vmax=extent
|
207 |
)
|
|
|
|
|
208 |
cbar = plt.colorbar(
|
209 |
cax,
|
210 |
orientation='horizontal',
|
211 |
+
pad=0.25,
|
212 |
+
aspect=40,
|
213 |
+
shrink=0.8
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
214 |
)
|
215 |
+
cbar.ax.tick_params(labelsize=8)
|
216 |
+
cbar.set_label('SHAP Contribution', fontsize=9, labelpad=5)
|
217 |
|
|
|
218 |
ax.set_yticks([])
|
219 |
ax.set_xlabel('Position in Sequence', fontsize=10)
|
220 |
ax.set_title(f"{title}{subtitle}", pad=10)
|
221 |
+
plt.subplots_adjust(bottom=0.25, left=0.05, right=0.95)
|
|
|
|
|
|
|
|
|
|
|
|
|
222 |
|
223 |
return fig
|
224 |
|
225 |
def create_importance_bar_plot(shap_values, kmers, top_k=10):
|
|
|
226 |
plt.rcParams.update({'font.size': 10})
|
227 |
fig = plt.figure(figsize=(10, 5))
|
228 |
|
|
|
229 |
indices = np.argsort(np.abs(shap_values))[-top_k:]
|
230 |
values = shap_values[indices]
|
231 |
features = [kmers[i] for i in indices]
|
232 |
|
|
|
233 |
colors = ['#99ccff' if v < 0 else '#ff9999' for v in values]
|
|
|
234 |
plt.barh(range(len(values)), values, color=colors)
|
235 |
plt.yticks(range(len(values)), features)
|
236 |
plt.xlabel('SHAP Value (impact on model output)')
|
|
|
240 |
return fig
|
241 |
|
242 |
def plot_shap_histogram(shap_array, title="SHAP Distribution in Region"):
|
|
|
|
|
|
|
243 |
fig, ax = plt.subplots(figsize=(6, 4))
|
244 |
ax.hist(shap_array, bins=30, color='gray', edgecolor='black')
|
245 |
ax.axvline(0, color='red', linestyle='--', label='0.0')
|
|
|
251 |
return fig
|
252 |
|
253 |
def compute_gc_content(sequence):
|
|
|
254 |
if not sequence:
|
255 |
return 0
|
256 |
gc_count = sequence.count('G') + sequence.count('C')
|
|
|
260 |
# 7. SEQUENCE ANALYSIS FUNCTIONS
|
261 |
###############################################################################
|
262 |
|
263 |
+
# Set up device and load the model once globally
|
264 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
265 |
+
model = VirusClassifier(256)
|
266 |
+
model.load_state_dict(torch.load("model.pt", map_location=device))
|
267 |
+
model.to(device)
|
268 |
+
model.eval()
|
269 |
+
|
270 |
+
KMERS_4 = [''.join(p) for p in product("ACGT", repeat=4)]
|
271 |
+
|
272 |
def analyze_sequence(file_path, top_k=10, fasta_text="", window_size=500):
|
273 |
"""
|
274 |
Analyze a virus sequence from a FASTA file or text input.
|
275 |
Returns (results_text, kmer_plot, heatmap_plot, state_dict, header)
|
276 |
"""
|
277 |
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
278 |
if file_path:
|
279 |
with open(file_path, 'r') as f:
|
280 |
fasta_text = f.read()
|
281 |
|
282 |
if not fasta_text.strip():
|
283 |
return ("Error: No sequence provided", None, None, {}, "")
|
284 |
+
|
|
|
285 |
sequences = parse_fasta(fasta_text)
|
286 |
if not sequences:
|
287 |
return ("Error: No valid FASTA sequences found", None, None, {}, "")
|
|
|
|
|
288 |
|
289 |
+
header, sequence = sequences[0]
|
290 |
+
|
291 |
+
x = sequence_to_kmer_vector(sequence, k=4)
|
292 |
+
x_tensor = torch.tensor(x).float().unsqueeze(0).to(device)
|
293 |
|
|
|
294 |
with torch.no_grad():
|
295 |
output = model(x_tensor)
|
296 |
probs = torch.softmax(output, dim=1)
|
|
|
297 |
pred_human = probs[0, 1].item()
|
|
|
|
|
|
|
298 |
|
299 |
+
classification = "Human" if pred_human > 0.5 else "Non-human"
|
300 |
+
|
301 |
+
shap_values, baseline_prob = calculate_shap_values(model, x_tensor)
|
302 |
+
|
303 |
+
shap_means = compute_positionwise_scores(sequence, shap_values, k=4)
|
304 |
+
|
305 |
start_max, end_max, avg_max = find_extreme_subregion(shap_means, window_size, mode="max")
|
306 |
start_min, end_min, avg_min = find_extreme_subregion(shap_means, window_size, mode="min")
|
307 |
|
|
|
|
|
308 |
results = (
|
309 |
f"Classification: {classification} "
|
310 |
f"(probability of human = {pred_human:.3f})\n\n"
|
311 |
f"Sequence length: {len(sequence):,} bases\n"
|
312 |
f"Overall GC content: {compute_gc_content(sequence):.1f}%\n\n"
|
313 |
+
f"Most human-like {window_size} bp region:\n"
|
314 |
f"Position {start_max:,} to {end_max:,}\n"
|
315 |
f"Average SHAP: {avg_max:.4f}\n"
|
316 |
f"GC content: {compute_gc_content(sequence[start_max:end_max]):.1f}%\n\n"
|
317 |
+
f"Least human-like {window_size} bp region:\n"
|
318 |
f"Position {start_min:,} to {end_min:,}\n"
|
319 |
f"Average SHAP: {avg_min:.4f}\n"
|
320 |
f"GC content: {compute_gc_content(sequence[start_min:end_min]):.1f}%"
|
321 |
)
|
322 |
|
323 |
+
kmer_fig = create_importance_bar_plot(shap_values, KMERS_4, top_k=top_k)
|
|
|
324 |
kmer_img = fig_to_image(kmer_fig)
|
325 |
|
|
|
326 |
heatmap_fig = plot_linear_heatmap(shap_means)
|
327 |
heatmap_img = fig_to_image(heatmap_fig)
|
328 |
|
|
|
329 |
state = {
|
330 |
"seq": sequence,
|
331 |
"shap_means": shap_means
|
|
|
334 |
return results, kmer_img, heatmap_img, state, header
|
335 |
|
336 |
except Exception as e:
|
337 |
+
return (f"Error analyzing sequence: {str(e)}", None, None, {}, "")
|
338 |
+
|
339 |
+
###############################################################################
|
340 |
+
# 8. SUBREGION ANALYSIS FUNCTION
|
341 |
+
###############################################################################
|
342 |
|
343 |
def analyze_subregion(state, header, region_start, region_end):
|
|
|
|
|
|
|
|
|
344 |
if not state or "seq" not in state or "shap_means" not in state:
|
345 |
return ("No sequence data found. Please run Step 1 first.", None, None)
|
346 |
|
347 |
seq = state["seq"]
|
348 |
shap_means = state["shap_means"]
|
349 |
|
|
|
350 |
region_start = int(region_start)
|
351 |
region_end = int(region_end)
|
352 |
|
|
|
355 |
if region_end <= region_start:
|
356 |
return ("Invalid region range. End must be > Start.", None, None)
|
357 |
|
|
|
358 |
region_seq = seq[region_start:region_end]
|
359 |
region_shap = shap_means[region_start:region_end]
|
360 |
|
|
|
361 |
gc_percent = compute_gc_content(region_seq)
|
362 |
avg_shap = float(np.mean(region_shap))
|
363 |
|
|
|
364 |
positive_fraction = np.mean(region_shap > 0)
|
365 |
negative_fraction = np.mean(region_shap < 0)
|
366 |
|
|
|
367 |
if avg_shap > 0.05:
|
368 |
region_classification = "Likely pushing toward human"
|
369 |
elif avg_shap < -0.05:
|
|
|
381 |
f"Subregion interpretation: {region_classification}\n"
|
382 |
)
|
383 |
|
|
|
384 |
heatmap_fig = plot_linear_heatmap(
|
385 |
shap_means,
|
386 |
title="Subregion SHAP",
|
|
|
389 |
)
|
390 |
heatmap_img = fig_to_image(heatmap_fig)
|
391 |
|
|
|
392 |
hist_fig = plot_shap_histogram(region_shap, title="SHAP Distribution in Subregion")
|
393 |
hist_img = fig_to_image(hist_fig)
|
394 |
|
395 |
return (region_info, heatmap_img, hist_img)
|
396 |
|
397 |
###############################################################################
|
398 |
+
# 9. COMPARISON ANALYSIS FUNCTIONS
|
399 |
###############################################################################
|
400 |
|
401 |
def normalize_shap_lengths(shap1, shap2, num_points=1000):
|
|
|
|
|
|
|
|
|
|
|
402 |
x1 = np.linspace(0, 1, len(shap1))
|
403 |
x2 = np.linspace(0, 1, len(shap2))
|
404 |
|
|
|
405 |
f1 = interp1d(x1, shap1, kind='linear')
|
406 |
f2 = interp1d(x2, shap2, kind='linear')
|
407 |
|
|
|
408 |
x_new = np.linspace(0, 1, num_points)
|
409 |
|
|
|
410 |
shap1_norm = f1(x_new)
|
411 |
shap2_norm = f2(x_new)
|
412 |
|
413 |
return shap1_norm, shap2_norm
|
414 |
|
415 |
def compute_shap_difference(shap1_norm, shap2_norm):
|
|
|
|
|
|
|
|
|
416 |
return shap2_norm - shap1_norm
|
417 |
|
418 |
def plot_comparative_heatmap(shap_diff, title="SHAP Difference Heatmap"):
|
|
|
|
|
|
|
|
|
|
|
419 |
heatmap_data = shap_diff.reshape(1, -1)
|
|
|
|
|
420 |
extent = max(abs(np.min(shap_diff)), abs(np.max(shap_diff)))
|
421 |
+
cmap = get_zero_centered_cmap()
|
422 |
|
|
|
423 |
fig, ax = plt.subplots(figsize=(12, 1.8))
|
|
|
|
|
|
|
|
|
|
|
424 |
cax = ax.imshow(
|
425 |
heatmap_data,
|
426 |
aspect='auto',
|
427 |
+
cmap=cmap,
|
428 |
vmin=-extent,
|
429 |
+
vmax=extent
|
430 |
)
|
|
|
|
|
431 |
cbar = plt.colorbar(
|
432 |
cax,
|
433 |
orientation='horizontal',
|
|
|
435 |
aspect=40,
|
436 |
shrink=0.8
|
437 |
)
|
|
|
|
|
438 |
cbar.ax.tick_params(labelsize=8)
|
439 |
+
cbar.set_label('SHAP Difference (Seq2 - Seq1)', fontsize=9, labelpad=5)
|
|
|
|
|
|
|
|
|
440 |
|
|
|
441 |
ax.set_yticks([])
|
442 |
ax.set_xlabel('Normalized Position (0-100%)', fontsize=10)
|
443 |
ax.set_title(title, pad=10)
|
444 |
+
plt.subplots_adjust(bottom=0.25, left=0.05, right=0.95)
|
|
|
|
|
|
|
|
|
|
|
445 |
|
446 |
return fig
|
447 |
|
448 |
def analyze_sequence_comparison(file1, file2, fasta1="", fasta2=""):
|
449 |
+
results1 = analyze_sequence(file1, top_k=10, fasta_text=fasta1, window_size=500)
|
|
|
|
|
|
|
|
|
|
|
450 |
if isinstance(results1[0], str) and "Error" in results1[0]:
|
451 |
return (f"Error in sequence 1: {results1[0]}", None, None)
|
452 |
|
453 |
+
results2 = analyze_sequence(file2, top_k=10, fasta_text=fasta2, window_size=500)
|
|
|
454 |
if isinstance(results2[0], str) and "Error" in results2[0]:
|
455 |
return (f"Error in sequence 2: {results2[0]}", None, None)
|
456 |
|
|
|
457 |
shap1 = results1[3]["shap_means"]
|
458 |
shap2 = results2[3]["shap_means"]
|
459 |
|
|
|
460 |
shap1_norm, shap2_norm = normalize_shap_lengths(shap1, shap2)
|
|
|
|
|
461 |
shap_diff = compute_shap_difference(shap1_norm, shap2_norm)
|
462 |
|
|
|
463 |
avg_diff = np.mean(shap_diff)
|
464 |
std_diff = np.std(shap_diff)
|
465 |
max_diff = np.max(shap_diff)
|
466 |
min_diff = np.min(shap_diff)
|
467 |
|
468 |
+
threshold = 0.05
|
|
|
469 |
substantial_diffs = np.abs(shap_diff) > threshold
|
470 |
frac_different = np.mean(substantial_diffs)
|
471 |
|
|
|
472 |
classification1 = results1[0].split('Classification: ')[1].split('\n')[0].strip()
|
473 |
classification2 = results2[0].split('Classification: ')[1].split('\n')[0].strip()
|
474 |
|
|
|
475 |
len1_formatted = "{:,}".format(len(shap1))
|
476 |
len2_formatted = "{:,}".format(len(shap2))
|
477 |
frac_formatted = "{:.2%}".format(frac_different)
|
478 |
|
|
|
479 |
comparison_text = (
|
480 |
"Sequence Comparison Results:\n"
|
481 |
f"Sequence 1: {results1[4]}\n"
|
|
|
495 |
"Negative values (blue) indicate regions where Sequence 1 is more 'human-like'"
|
496 |
)
|
497 |
|
|
|
498 |
heatmap_fig = plot_comparative_heatmap(shap_diff)
|
499 |
heatmap_img = fig_to_image(heatmap_fig)
|
500 |
|
501 |
+
hist_fig = plot_shap_histogram(shap_diff, title="Distribution of SHAP Differences")
|
|
|
|
|
|
|
|
|
502 |
hist_img = fig_to_image(hist_fig)
|
503 |
|
504 |
return comparison_text, heatmap_img, hist_img
|
505 |
|
506 |
###############################################################################
|
507 |
+
# 10. BUILD GRADIO INTERFACE
|
508 |
###############################################################################
|
509 |
|
510 |
css = """
|
|
|
535 |
placeholder=">sequence_name\nACGTACGT...",
|
536 |
lines=5
|
537 |
)
|
538 |
+
top_k_slider = gr.Slider(
|
539 |
minimum=5,
|
540 |
maximum=30,
|
541 |
value=10,
|
542 |
step=1,
|
543 |
label="Number of top k-mers to display"
|
544 |
)
|
545 |
+
win_size_slider = gr.Slider(
|
546 |
minimum=100,
|
547 |
maximum=5000,
|
548 |
value=500,
|
|
|
563 |
|
564 |
analyze_btn.click(
|
565 |
analyze_sequence,
|
566 |
+
inputs=[file_input, top_k_slider, text_input, win_size_slider],
|
567 |
outputs=[results_box, kmer_img, genome_img, seq_state, header_state]
|
568 |
)
|
569 |
|
|
|
666 |
- Statistical summary of differences
|
667 |
""")
|
668 |
|
|
|
|
|
|
|
|
|
669 |
if __name__ == "__main__":
|
|
|
670 |
plt.style.use('default')
|
671 |
plt.rcParams['figure.figsize'] = [10, 6]
|
672 |
plt.rcParams['figure.dpi'] = 100
|
673 |
plt.rcParams['font.size'] = 10
|
674 |
|
|
|
675 |
iface.launch(
|
676 |
+
share=False,
|
677 |
+
server_name="0.0.0.0",
|
678 |
+
server_port=7860,
|
679 |
+
show_api=False,
|
680 |
+
debug=False
|
681 |
)
|