File size: 13,967 Bytes
6d1bbc7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
"""Tests for CT feature encoding (ct_features.py).

5 test classes:
  TestEncodeDrugFeatures: 6 tests
  TestEncodeConditionFeatures: 2 tests
  TestEncodeTrialFeatures: 5 tests
  TestBuildFeatureMatrix: 4 tests
  TestGNNTabFeatures: 2 tests
"""

import numpy as np
import pandas as pd
import pytest

from negbiodb_ct.ct_features import (
    BLINDING_KEYWORDS,
    CONDITION_DIM,
    DRUG_FP_DIM,
    DRUG_TAB_DIM,
    M2_TRIAL_DIM,
    MOLECULAR_TYPES,
    PHASE_ORDER,
    SPONSOR_TYPES,
    TOTAL_M1_DIM,
    TOTAL_M2_DIM,
    TRIAL_PHASES,
    _encode_blinding,
    _one_hot,
    build_feature_matrix,
    build_gnn_tab_features,
    build_mlp_features,
    build_xgboost_features,
    encode_condition_features,
    encode_drug_features,
    encode_trial_features,
)

# Real SMILES for testing
ASPIRIN_SMILES = "CC(=O)Oc1ccccc1C(=O)O"
CAFFEINE_SMILES = "Cn1c(=O)c2c(ncn2C)n(C)c1=O"


def _make_drug_df(
    n: int = 3,
    *,
    with_smiles: bool = True,
    with_all_cols: bool = True,
) -> pd.DataFrame:
    """Build a minimal drug DataFrame for testing."""
    data: dict = {
        "smiles": [ASPIRIN_SMILES, CAFFEINE_SMILES, None][:n] if with_smiles else [None] * n,
        "molecular_type": ["small_molecule", "small_molecule", "monoclonal_antibody"][:n],
    }
    if with_all_cols:
        data["target_count"] = [2, 0, 5][:n]
        data["intervention_degree"] = [10, 1, 50][:n]
        data["highest_phase_reached"] = ["phase_3", "phase_1", None][:n]
        data["condition_degree"] = [5, 1, 20][:n]
    return pd.DataFrame(data)


def _make_m2_df(n: int = 3) -> pd.DataFrame:
    """Build a minimal M2 DataFrame with trial features."""
    df = _make_drug_df(n, with_smiles=True, with_all_cols=True)
    df["trial_phase"] = ["phase_2", "phase_3", "not_applicable"][:n]
    df["blinding"] = [
        "Double (Participant, Investigator)",
        "None (Open Label)",
        None,
    ][:n]
    df["sponsor_type"] = ["industry", "academic", None][:n]
    df["randomized"] = [1, 0, None][:n]
    df["enrollment_actual"] = [500, 30, None][:n]
    df["failure_category"] = ["efficacy", "safety", "enrollment"][:n]
    df["failure_category_int"] = [0, 4, 1][:n]
    return df


# ============================================================================
# TestEncodeDrugFeatures
# ============================================================================


class TestEncodeDrugFeatures:
    """Test encode_drug_features with various inputs."""

    def test_dims_with_fp(self):
        """With FP: output should be (N, DRUG_FP_DIM + DRUG_TAB_DIM)."""
        df = _make_drug_df(3)
        result = encode_drug_features(df, include_fp=True)
        assert result.shape == (3, DRUG_FP_DIM + DRUG_TAB_DIM)
        # 1030 + 13 = 1043
        assert result.shape[1] == 1043

    def test_dims_without_fp(self):
        """Without FP: output should be (N, DRUG_TAB_DIM)."""
        df = _make_drug_df(3)
        result = encode_drug_features(df, include_fp=False)
        assert result.shape == (3, DRUG_TAB_DIM)
        assert result.shape[1] == 13

    def test_nan_for_missing_smiles(self):
        """Row 2 has None SMILES → FP columns should be NaN."""
        df = _make_drug_df(3)
        result = encode_drug_features(df, include_fp=True)
        # Row 2 (no SMILES) should have NaN in FP region
        assert np.all(np.isnan(result[2, :DRUG_FP_DIM]))
        # Row 0 (aspirin) should NOT have NaN in FP region
        assert not np.any(np.isnan(result[0, :DRUG_FP_DIM]))

    def test_fp_nonzero_for_valid_smiles(self):
        """Valid SMILES should produce nonzero FP bits."""
        df = _make_drug_df(1)
        result = encode_drug_features(df, include_fp=True)
        # Aspirin FP should have at least some bits set
        fp_region = result[0, :1024]
        assert np.sum(fp_region) > 0

    def test_molecular_type_one_hot(self):
        """Molecular type should use 10-dim one-hot (9 types + unknown)."""
        df = _make_drug_df(3, with_smiles=False)
        result = encode_drug_features(df, include_fp=False)
        # First 10 columns are mol_type one-hot
        mol_type_region = result[:, :len(MOLECULAR_TYPES) + 1]
        assert mol_type_region.shape[1] == 10
        # Each row should sum to 1 (one-hot)
        for i in range(3):
            assert mol_type_region[i].sum() == 1.0
        # Row 0: small_molecule → index 0
        assert mol_type_region[0, 0] == 1.0
        # Row 2: monoclonal_antibody → index 1
        assert mol_type_region[2, 1] == 1.0

    def test_coalesce_missing_columns(self):
        """Missing target_count/degree/phase columns → defaults, no error."""
        df = _make_drug_df(2, with_all_cols=False)
        result = encode_drug_features(df, include_fp=False)
        assert result.shape == (2, DRUG_TAB_DIM)
        # Should not contain NaN (all defaults applied)
        assert not np.any(np.isnan(result))

    def test_deterministic(self):
        """Same input → same output."""
        df = _make_drug_df(3)
        r1 = encode_drug_features(df, include_fp=True)
        r2 = encode_drug_features(df, include_fp=True)
        np.testing.assert_array_equal(r1, r2)


# ============================================================================
# TestEncodeConditionFeatures
# ============================================================================


class TestEncodeConditionFeatures:
    """Test encode_condition_features."""

    def test_dims(self):
        """Output should be (N, 1)."""
        df = pd.DataFrame({"condition_degree": [5, 1, 20]})
        result = encode_condition_features(df)
        assert result.shape == (3, CONDITION_DIM)
        assert result.shape[1] == 1

    def test_log_transform(self):
        """Values should be log1p-transformed."""
        df = pd.DataFrame({"condition_degree": [0, 1, 100]})
        result = encode_condition_features(df)
        np.testing.assert_allclose(result[0, 0], np.log1p(0), rtol=1e-5)
        np.testing.assert_allclose(result[1, 0], np.log1p(1), rtol=1e-5)
        np.testing.assert_allclose(result[2, 0], np.log1p(100), rtol=1e-5)

    def test_missing_column_default(self):
        """Missing condition_degree column → default to 1."""
        df = pd.DataFrame({"other_col": [1, 2]})
        result = encode_condition_features(df)
        assert result.shape == (2, 1)
        np.testing.assert_allclose(result[0, 0], np.log1p(1), rtol=1e-5)


# ============================================================================
# TestEncodeTrialFeatures
# ============================================================================


class TestEncodeTrialFeatures:
    """Test encode_trial_features for M2."""

    def test_dims(self):
        """Output should be (N, 22)."""
        df = _make_m2_df(3)
        result = encode_trial_features(df)
        assert result.shape == (3, M2_TRIAL_DIM)
        assert result.shape[1] == 22

    def test_trial_phase_one_hot(self):
        """Trial phase should be 9-dim one-hot."""
        df = _make_m2_df(3)
        result = encode_trial_features(df)
        # First 9 columns are trial_phase one-hot
        tp_region = result[:, :len(TRIAL_PHASES) + 1]
        assert tp_region.shape[1] == 9
        # Each row sums to 1
        for i in range(3):
            assert tp_region[i].sum() == 1.0
        # Row 0: phase_2 → index 3
        assert tp_region[0, TRIAL_PHASES.index("phase_2")] == 1.0
        # Row 1: phase_3 → index 5
        assert tp_region[1, TRIAL_PHASES.index("phase_3")] == 1.0

    def test_blinding_case_insensitive(self):
        """Blinding should match case-insensitively via substring."""
        df = _make_m2_df(3)
        result = encode_trial_features(df)
        # Blinding is at columns 9:15 (after trial_phase 9-dim)
        bl_start = len(TRIAL_PHASES) + 1  # 9
        bl_end = bl_start + len(BLINDING_KEYWORDS) + 1  # 9 + 6 = 15
        bl_region = result[:, bl_start:bl_end]
        # Row 0: "Double (Participant, Investigator)" → "double" match → index 2
        assert bl_region[0, BLINDING_KEYWORDS.index("double")] == 1.0
        # Row 1: "None (Open Label)" → "none" match → index 0
        assert bl_region[1, BLINDING_KEYWORDS.index("none")] == 1.0
        # Row 2: None → unknown bucket (index 5)
        assert bl_region[2, -1] == 1.0

    def test_sponsor_one_hot(self):
        """Sponsor type should use 5-dim one-hot (4 types + unknown)."""
        df = _make_m2_df(3)
        result = encode_trial_features(df)
        sp_start = (len(TRIAL_PHASES) + 1) + (len(BLINDING_KEYWORDS) + 1)  # 9 + 6 = 15
        sp_end = sp_start + len(SPONSOR_TYPES) + 1  # 15 + 5 = 20
        sp_region = result[:, sp_start:sp_end]
        # Row 0: industry → index 0
        assert sp_region[0, SPONSOR_TYPES.index("industry")] == 1.0
        # Row 1: academic → index 1
        assert sp_region[1, SPONSOR_TYPES.index("academic")] == 1.0
        # Row 2: None → unknown bucket
        assert sp_region[2, -1] == 1.0

    def test_enrollment_log1p(self):
        """Enrollment should be log1p-transformed."""
        df = _make_m2_df(3)
        result = encode_trial_features(df)
        # enrollment is the last column (index 21)
        np.testing.assert_allclose(result[0, -1], np.log1p(500), rtol=1e-5)
        np.testing.assert_allclose(result[1, -1], np.log1p(30), rtol=1e-5)
        # Row 2: None → log1p(0) = 0
        np.testing.assert_allclose(result[2, -1], 0.0, atol=1e-6)


# ============================================================================
# TestBuildFeatureMatrix
# ============================================================================


class TestBuildFeatureMatrix:
    """Test composite build functions."""

    def test_m1_dims(self):
        """M1 with FP: (N, 1044)."""
        df = _make_drug_df(3)
        result = build_feature_matrix(df, task="m1", include_fp=True)
        assert result.shape == (3, TOTAL_M1_DIM)
        assert result.shape[1] == 1044

    def test_m2_dims(self):
        """M2 with FP: (N, 1066)."""
        df = _make_m2_df(3)
        result = build_feature_matrix(df, task="m2", include_fp=True)
        assert result.shape == (3, TOTAL_M2_DIM)
        assert result.shape[1] == 1066

    def test_xgboost_nan_preserved(self):
        """XGBoost features should preserve NaN for missing SMILES."""
        df = _make_drug_df(3)
        result = build_xgboost_features(df, task="m1")
        # Row 2 has no SMILES → NaN in FP region
        assert np.any(np.isnan(result[2, :DRUG_FP_DIM]))

    def test_mlp_zero_padded(self):
        """MLP features should replace NaN with 0.0."""
        df = _make_drug_df(3)
        result = build_mlp_features(df, task="m1")
        assert not np.any(np.isnan(result))
        # Row 2 FP region should be all zeros (zero-padded)
        assert np.all(result[2, :DRUG_FP_DIM] == 0.0)


# ============================================================================
# TestGNNTabFeatures
# ============================================================================


class TestGNNTabFeatures:
    """Test GNN tabular feature builder (no FP)."""

    def test_m1_dims(self):
        """GNN tab M1: (N, DRUG_TAB_DIM + CONDITION_DIM) = (N, 14)."""
        df = _make_drug_df(3)
        result = build_gnn_tab_features(df, task="m1")
        expected_dim = DRUG_TAB_DIM + CONDITION_DIM
        assert result.shape == (3, expected_dim)
        assert result.shape[1] == 14

    def test_m2_dims(self):
        """GNN tab M2: (N, DRUG_TAB_DIM + CONDITION_DIM + M2_TRIAL_DIM) = (N, 36)."""
        df = _make_m2_df(3)
        result = build_gnn_tab_features(df, task="m2")
        expected_dim = DRUG_TAB_DIM + CONDITION_DIM + M2_TRIAL_DIM
        assert result.shape == (3, expected_dim)
        assert result.shape[1] == 36

    def test_no_nan_in_output(self):
        """GNN tab features should have no NaN (zero-padded)."""
        df = _make_drug_df(3)
        result = build_gnn_tab_features(df, task="m1")
        assert not np.any(np.isnan(result))


# ============================================================================
# Additional edge-case tests
# ============================================================================


class TestHelpers:
    """Test helper functions directly."""

    def test_one_hot_known_value(self):
        """Known value → correct index set."""
        result = _one_hot("phase_2", TRIAL_PHASES)
        assert len(result) == len(TRIAL_PHASES) + 1
        assert result[TRIAL_PHASES.index("phase_2")] == 1
        assert sum(result) == 1

    def test_one_hot_unknown_value(self):
        """Unknown value → last bucket."""
        result = _one_hot("bogus", TRIAL_PHASES)
        assert result[-1] == 1
        assert sum(result) == 1

    def test_one_hot_none(self):
        """None → unknown bucket."""
        result = _one_hot(None, TRIAL_PHASES)
        assert result[-1] == 1

    def test_encode_blinding_mixed_case(self):
        """Case variations should all match."""
        assert _encode_blinding("DOUBLE")[BLINDING_KEYWORDS.index("double")] == 1
        assert _encode_blinding("Double (Participant)")[BLINDING_KEYWORDS.index("double")] == 1
        assert _encode_blinding("single blind")[BLINDING_KEYWORDS.index("single")] == 1

    def test_encode_blinding_none(self):
        """None blinding → unknown bucket."""
        result = _encode_blinding(None)
        assert result[-1] == 1
        assert sum(result) == 1

    def test_dimension_constants_consistent(self):
        """Verify dimension constants add up correctly."""
        assert DRUG_FP_DIM == 1024 + 6  # FP_NBITS + N_MOL_PROPS
        assert DRUG_TAB_DIM == 10 + 1 + 1 + 1  # mol_type + tc + deg + phase
        assert CONDITION_DIM == 1
        assert M2_TRIAL_DIM == 9 + 6 + 5 + 1 + 1  # tp + bl + sp + rand + enr
        assert TOTAL_M1_DIM == DRUG_FP_DIM + DRUG_TAB_DIM + CONDITION_DIM
        assert TOTAL_M2_DIM == TOTAL_M1_DIM + M2_TRIAL_DIM