File size: 11,800 Bytes
81d44ef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
9c47d54
81d44ef
9c47d54
81d44ef
 
9c47d54
 
 
 
 
81d44ef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
"""SHAP-guided target site alignment visualization helpers."""

from __future__ import annotations

import io

import matplotlib.patches as patches
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image


COLOR_PALETTE = [
    '#034da1', '#3359a7', '#4c66ad', '#6174b3', '#7482b9',
    '#8590be', '#979ec4', '#a8adca', '#b9bbcf', '#c9cbd5',
    '#dadada', '#dac4c4', '#daaeae', '#da9999', '#da8383',
    '#da6d6d', '#d95757', '#d94141', '#d92c2c', '#d91616', '#d90000',
]

COMPLEMENT = {
    'A': 'T',
    'T': 'A',
    'C': 'G',
    'G': 'C',
    '-': '',
    'N': '',
}


def _plot_t(ax, base, left_edge, height, color):
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.4, base], width=0.2, height=height,
        facecolor=color, edgecolor=color, fill=True
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge, base + 0.85 * height], width=1.0, height=0.15 * height,
        facecolor=color, edgecolor=color, fill=True
    ))


def _plot_g(ax, base, left_edge, height, color):
    ax.add_patch(patches.Ellipse(
        xy=[left_edge + 0.65, base + 0.5 * height], width=1.3, height=height,
        facecolor=color, edgecolor=color
    ))
    ax.add_patch(patches.Ellipse(
        xy=[left_edge + 0.65, base + 0.5 * height], width=0.91, height=0.7 * height,
        facecolor='white', edgecolor='white'
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 1, base], width=1.0, height=height,
        facecolor='white', edgecolor='white', fill=True
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.825, base + 0.085 * height], width=0.174, height=0.415 * height,
        facecolor=color, edgecolor=color, fill=True
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.625, base + 0.35 * height], width=0.374, height=0.15 * height,
        facecolor=color, edgecolor=color, fill=True
    ))


def _plot_a(ax, base, left_edge, height, color):
    polygons = [
        np.array([[0.0, 0.0], [0.5, 1.0], [0.5, 0.8], [0.2, 0.0]]),
        np.array([[1.0, 0.0], [0.5, 1.0], [0.5, 0.8], [0.8, 0.0]]),
        np.array([[0.225, 0.45], [0.775, 0.45], [0.85, 0.3], [0.15, 0.3]]),
    ]
    scale = np.array([1, height])[None, :]
    offset = np.array([left_edge, base])[None, :]
    for polygon in polygons:
        ax.add_patch(patches.Polygon(polygon * scale + offset, facecolor=color, edgecolor=color))


def _plot_c(ax, base, left_edge, height, color):
    ax.add_patch(patches.Ellipse(
        xy=[left_edge + 0.65, base + 0.5 * height], width=1.3, height=height,
        facecolor=color, edgecolor=color
    ))
    ax.add_patch(patches.Ellipse(
        xy=[left_edge + 0.65, base + 0.5 * height], width=0.91, height=0.7 * height,
        facecolor='white', edgecolor='white'
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 1, base], width=1.0, height=height,
        facecolor='white', edgecolor='white', fill=True
    ))


def _plot_n(ax, base, left_edge, height, color):
    ax.add_patch(patches.Rectangle(
        xy=[left_edge, base], width=0.2, height=height,
        facecolor=color, edgecolor=color, fill=True
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.8, base], width=0.15, height=height,
        facecolor=color, edgecolor=color, fill=True, angle=45
    ))
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.8, base], width=0.2, height=height,
        facecolor=color, edgecolor=color, fill=True
    ))


def _plot_dash(ax, base, left_edge, height, color):
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.2, base + 0.3 * height], width=0.6, height=0.2 * height,
        facecolor=color, edgecolor=color, fill=True
    ))


def _plot_line(ax, base, left_edge, height, color):
    ax.add_patch(patches.Rectangle(
        xy=[left_edge + 0.45, base + 0.2 * height], width=0.1, height=0.6 * height,
        facecolor=color, edgecolor=color, fill=True
    ))


def _plot_dot(ax, base, left_edge, height, color):
    ax.add_patch(patches.Ellipse(
        xy=[left_edge + 0.5, base + 0.5 * height], width=0.3, height=0.2 * height,
        facecolor=color, edgecolor=color
    ))


BASE_PLOTTERS = {
    'A': _plot_a,
    'C': _plot_c,
    'T': _plot_t,
    'G': _plot_g,
    'N': _plot_n,
    '-': _plot_dash,
}


class _ShapAlignment:
    """Needleman-Wunsch-style traceback with SHAP-informed scores."""

    RIGHT = "right"
    DOWN = "down"
    DIAGONAL = "diag"

    def align(self, mirna_seq, target_seq, score_matrix,
              opening_percentile=99, elongating_percentile=90):
        processed_scores, opening_gap, elongating_gap = self._preprocess_score(
            score_matrix, opening_percentile, elongating_percentile
        )
        grid = self._forward_pass(mirna_seq, target_seq, processed_scores, opening_gap, elongating_gap)
        return self._backward_pass(mirna_seq, target_seq, grid, processed_scores)

    def _score(self, grid, i, j, gap_penalty, score):
        directions = [self.RIGHT, self.DIAGONAL, self.DOWN]
        values = [
            grid[i, j - 1][1] - gap_penalty,
            grid[i - 1, j - 1][1] + score,
            grid[i - 1, j][1] - gap_penalty,
        ]
        best_index = int(np.argmax(values))
        return directions[best_index], values[best_index]

    def _forward_pass(self, mirna_seq, target_seq, score_matrix, opening_gap, elongating_gap):
        target_bases = [""] + list(target_seq)
        mirna_bases = [""] + list(mirna_seq)

        grid = np.empty((len(target_bases), len(mirna_bases)), dtype=object)
        for i in range(len(target_bases)):
            grid[i, 0] = (self.DOWN, 0.0)
        for j in range(len(mirna_bases)):
            grid[0, j] = (self.RIGHT, 0.0)

        is_opening = False
        for i in range(1, len(target_bases)):
            for j in range(1, len(mirna_bases)):
                gap_penalty = opening_gap if is_opening else elongating_gap
                if i == len(target_bases) - 1 or j == len(mirna_bases) - 1:
                    gap_penalty = 0.0
                grid[i, j] = self._score(grid, i, j, gap_penalty, score_matrix[i, j])
                is_opening = grid[i, j][0] == self.DIAGONAL

        return grid

    def _backward_pass(self, mirna_seq, target_seq, grid, score_matrix):
        target_bases = [""] + list(target_seq)
        mirna_bases = [""] + list(mirna_seq)

        aligned_target = []
        aligned_mirna = []
        aligned_scores = []

        i = grid.shape[0] - 1
        j = grid.shape[1] - 1
        while i != 0 or j != 0:
            direction = grid[i, j][0]
            if direction == self.RIGHT:
                aligned_target.append("-")
                aligned_mirna.append(mirna_bases[j])
                aligned_scores.append(0.0)
                j -= 1
            elif direction == self.DOWN:
                aligned_target.append(target_bases[i])
                aligned_mirna.append("-")
                aligned_scores.append(0.0)
                i -= 1
            else:
                aligned_target.append(target_bases[i])
                aligned_mirna.append(mirna_bases[j])
                aligned_scores.append(score_matrix[i, j])
                i -= 1
                j -= 1

        return aligned_target, aligned_scores, aligned_mirna

    def _preprocess_score(self, score_matrix, opening_percentile, elongating_percentile):
        score_matrix = np.asarray(score_matrix, dtype=float)
        if score_matrix.size == 0:
            return np.zeros((1, 1), dtype=float), 0.0, 0.0

        max_absolute_value = float(np.max(np.abs(score_matrix)))
        if max_absolute_value > 0:
            score_matrix = score_matrix / max_absolute_value
        else:
            score_matrix = np.zeros_like(score_matrix)

        abs_values = np.abs(score_matrix.sum(axis=-1)).ravel()
        opening_gap = float(np.nanpercentile(abs_values, opening_percentile)) if abs_values.size else 0.0
        elongating_gap = float(np.nanpercentile(abs_values, elongating_percentile)) if abs_values.size else 0.0

        score_matrix = np.vstack([np.zeros(score_matrix.shape[1]), score_matrix])
        score_matrix = np.hstack([np.zeros((score_matrix.shape[0], 1)), score_matrix])

        return score_matrix, opening_gap, elongating_gap


def _scale_scores(alignment_scores):
    values = np.asarray(alignment_scores, dtype=float)
    if values.size == 0:
        return np.array([], dtype=int)

    max_absolute_value = float(np.max(np.abs(values)))
    if max_absolute_value == 0.0:
        return np.full(values.shape, 10, dtype=int)

    scaled = values / max_absolute_value
    scaled = scaled * 10 + 10
    return np.clip(np.rint(scaled).astype(int), 0, len(COLOR_PALETTE) - 1)


def compute_shap_alignment(mirna_seq, target_seq, shap_2d,
                           opening_percentile=99, elongating_percentile=90):
    """Align miRNA and target using SHAP attributions as pairwise scores."""
    shap_array = np.asarray(shap_2d, dtype=float)
    expected_shape = (len(mirna_seq), len(target_seq))
    if shap_array.shape != expected_shape:
        raise ValueError(
            f"Expected shap_2d shape {expected_shape}, got {tuple(shap_array.shape)}"
        )

    aligner = _ShapAlignment()
    aligned_target, aligned_scores, aligned_mirna = aligner.align(
        mirna_seq[::-1],
        target_seq,
        shap_array[::-1].T,
        opening_percentile=opening_percentile,
        elongating_percentile=elongating_percentile,
    )
    return aligned_target[::-1], aligned_scores[::-1], aligned_mirna[::-1]


def plot_alignment_image(mirna_seq, target_seq, shap_2d, arrows=True):
    """Render the SHAP-guided alignment as a PIL image."""
    aligned_target, aligned_scores, aligned_mirna = compute_shap_alignment(
        mirna_seq, target_seq, shap_2d
    )
    color_indices = _scale_scores(aligned_scores)

    fig_width = max(12, len(aligned_target) * 0.42)
    fig, ax = plt.subplots(figsize=(fig_width, 3.2))

    step = 1.2
    left_margin = 2.8
    alignment_width = len(aligned_target) * step
    ax.set_xlim(-left_margin, alignment_width + 0.6)
    ax.set_ylim(0, 4.8)

    ax.text(-left_margin + 0.2, 3.5, "Target", ha='left', va='center',
            fontsize=12, fontweight='bold')
    ax.text(-left_margin + 0.2, 1.1, "miRNA", ha='left', va='center',
            fontsize=12, fontweight='bold')

    if arrows:
        ax.add_patch(patches.Arrow(x=0, y=4.4, dx=2, dy=0, width=0.6, color=COLOR_PALETTE[10]))
        ax.add_patch(patches.Arrow(
            x=max(alignment_width - 0.2, 0), y=0.2, dx=-2, dy=0, width=0.6, color=COLOR_PALETTE[10]
        ))

    ax.text(-0.1, 4.45, "5'", ha='right', va='center', fontsize=10, fontweight='bold')
    ax.text(alignment_width + 0.15, 4.45, "3'", ha='left', va='center', fontsize=10, fontweight='bold')
    ax.text(-0.1, 0.15, "3'", ha='right', va='center', fontsize=10, fontweight='bold')
    ax.text(alignment_width + 0.15, 0.15, "5'", ha='left', va='center', fontsize=10, fontweight='bold')

    for i, (target_base, mirna_base, color_index) in enumerate(
        zip(aligned_target, aligned_mirna, color_indices)
    ):
        color = COLOR_PALETTE[int(color_index)]
        left_edge = step * i

        BASE_PLOTTERS.get(mirna_base, _plot_n)(ax, 0.6, left_edge, 1.0, color)
        if mirna_base == COMPLEMENT.get(target_base):
            _plot_line(ax, 1.7, left_edge, 1.0, color)
        elif mirna_base in "ACTG" and target_base in "ACTG":
            _plot_dot(ax, 1.7, left_edge, 1.0, color)
        BASE_PLOTTERS.get(target_base, _plot_n)(ax, 3.0, left_edge, 1.0, color)

    ax.axis('off')
    fig.tight_layout()

    buffer = io.BytesIO()
    fig.savefig(buffer, format='png', dpi=120, bbox_inches='tight')
    buffer.seek(0)
    image = Image.open(buffer)
    plt.close(fig)
    return image