File size: 11,006 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
"""Tests for GE LLM evaluation module (parsing + metrics)."""

import pytest

from negbiodb_depmap.llm_eval import (
    compute_all_ge_llm_metrics,
    evaluate_ge_l1,
    evaluate_ge_l2,
    evaluate_ge_l3,
    evaluate_ge_l4,
    parse_ge_l1_answer,
    parse_ge_l2_response,
    parse_ge_l3_judge_scores,
    parse_ge_l4_answer,
)


# ── L1 parsing ────────────────────────────────────────────────────────────


class TestParseL1:
    def test_single_letter(self):
        assert parse_ge_l1_answer("C") == "C"

    def test_lowercase(self):
        assert parse_ge_l1_answer("b") == "B"

    def test_answer_prefix(self):
        assert parse_ge_l1_answer("Answer: A") == "A"

    def test_parenthesized(self):
        assert parse_ge_l1_answer("(D)") == "D"

    def test_with_explanation(self):
        assert parse_ge_l1_answer("C\nBecause the gene is non-essential...") == "C"

    def test_empty(self):
        assert parse_ge_l1_answer("") is None

    def test_no_valid_letter(self):
        # Text without any A-D characters
        assert parse_ge_l1_answer("no help here with this question") is None

    def test_embedded_letter(self):
        assert parse_ge_l1_answer("I think B is correct") == "B"


class TestEvaluateL1:
    def test_perfect_accuracy(self):
        preds = ["A", "B", "C", "D"]
        gold = ["A", "B", "C", "D"]
        result = evaluate_ge_l1(preds, gold)
        assert result["accuracy"] == 1.0
        assert result["valid_rate"] == 1.0

    def test_zero_accuracy(self):
        preds = ["B", "C", "D", "A"]
        gold = ["A", "B", "C", "D"]
        result = evaluate_ge_l1(preds, gold)
        assert result["accuracy"] == 0.0

    def test_invalid_responses(self):
        # "nonsense" has no A-D letters, so it's truly unparseable
        preds = ["nonsense", "A"]
        gold = ["A", "A"]
        result = evaluate_ge_l1(preds, gold)
        assert result["n_valid"] == 1
        assert result["valid_rate"] == 0.5

    def test_empty_predictions(self):
        result = evaluate_ge_l1([], [])
        assert result["accuracy"] == 0.0


# ── L2 parsing ────────────────────────────────────────────────────────────


class TestParseL2:
    def test_valid_json(self):
        raw = '{"genes": [{"gene_name": "TP53"}], "total_genes_mentioned": 1, "screen_type": "CRISPR"}'
        result = parse_ge_l2_response(raw)
        assert result is not None
        assert result["total_genes_mentioned"] == 1

    def test_markdown_json(self):
        raw = '```json\n{"genes": [], "total_genes_mentioned": 0, "screen_type": "RNAi"}\n```'
        result = parse_ge_l2_response(raw)
        assert result is not None

    def test_json_with_text(self):
        raw = 'Here is the extraction:\n{"genes": [], "total_genes_mentioned": 0, "screen_type": "CRISPR"}'
        result = parse_ge_l2_response(raw)
        assert result is not None

    def test_invalid_json(self):
        assert parse_ge_l2_response("not json at all") is None

    def test_empty(self):
        assert parse_ge_l2_response("") is None


class TestEvaluateL2:
    def test_perfect_parse(self):
        preds = ['{"genes": [], "total_genes_mentioned": 0, "screen_type": "CRISPR"}']
        gold = [{"genes": [], "total_genes_mentioned": 0, "screen_type": "CRISPR"}]
        result = evaluate_ge_l2(preds, gold)
        assert result["parse_rate"] == 1.0
        assert result["schema_compliance"] == 1.0

    def test_missing_fields(self):
        preds = ['{"genes": []}']
        gold = [{"genes": [], "total_genes_mentioned": 0, "screen_type": "CRISPR"}]
        result = evaluate_ge_l2(preds, gold)
        assert result["schema_compliance"] == 0.0

    def test_unparseable(self):
        preds = ["invalid"]
        gold = [{"genes": []}]
        result = evaluate_ge_l2(preds, gold)
        assert result["parse_rate"] == 0.0

    def test_essentiality_accuracy_correct(self):
        preds = ['{"genes": [{"gene_name": "TP53", "essentiality_status": "non-essential"}], "total_genes_mentioned": 1, "screen_type": "CRISPR"}']
        gold = [{"genes": [{"gene_name": "TP53", "essentiality_status": "non-essential"}], "total_genes_mentioned": 1, "screen_type": "CRISPR"}]
        result = evaluate_ge_l2(preds, gold)
        assert result["essentiality_accuracy"] == 1.0
        assert result["essentiality_n"] == 1

    def test_essentiality_accuracy_wrong(self):
        preds = ['{"genes": [{"gene_name": "TP53", "essentiality_status": "essential"}], "total_genes_mentioned": 1, "screen_type": "CRISPR"}']
        gold = [{"genes": [{"gene_name": "TP53", "essentiality_status": "non-essential"}], "total_genes_mentioned": 1, "screen_type": "CRISPR"}]
        result = evaluate_ge_l2(preds, gold)
        assert result["essentiality_accuracy"] == 0.0

    def test_legacy_essentiality_findings_key(self):
        # gold uses old schema key 'essentiality_findings' — should still work
        preds = ['{"genes": [{"gene_name": "BRCA1", "essentiality_status": "non-essential"}], "total_genes_mentioned": 1, "screen_type": "CRISPR"}']
        gold = [{"essentiality_findings": [{"gene_name": "BRCA1", "essentiality_status": "non-essential"}], "total_gene_count": 1, "screen_type": "CRISPR"}]
        result = evaluate_ge_l2(preds, gold)
        assert result["essentiality_accuracy"] == 1.0


# ── L3 parsing ────────────────────────────────────────────────────────────


class TestParseL3:
    def test_standard_format(self):
        raw = """biological_plausibility: 4
pathway_reasoning: 3
context_specificity: 5
mechanistic_depth: 4"""
        result = parse_ge_l3_judge_scores(raw)
        assert result is not None
        assert result["biological_plausibility"] == 4.0
        assert result["context_specificity"] == 5.0

    def test_json_format(self):
        raw = '{"biological_plausibility": 3, "pathway_reasoning": 4, "context_specificity": 3, "mechanistic_depth": 2}'
        result = parse_ge_l3_judge_scores(raw)
        assert result is not None
        assert result["pathway_reasoning"] == 4.0

    def test_empty(self):
        assert parse_ge_l3_judge_scores("") is None


class TestEvaluateL3:
    def test_basic_evaluation(self):
        judge_outputs = [
            "biological_plausibility: 4\npathway_reasoning: 3\ncontext_specificity: 5\nmechanistic_depth: 4",
            "biological_plausibility: 3\npathway_reasoning: 4\ncontext_specificity: 3\nmechanistic_depth: 3",
        ]
        result = evaluate_ge_l3(judge_outputs)
        assert result["n_parsed"] == 2
        assert result["biological_plausibility_mean"] == 3.5
        assert result["overall_mean"] > 0

    def test_no_parseable(self):
        result = evaluate_ge_l3(["invalid", "garbage"])
        assert result["n_parsed"] == 0
        assert result["overall_mean"] == 0.0


# ── L4 parsing ────────────────────────────────────────────────────────────


class TestParseL4:
    def test_tested(self):
        assert parse_ge_l4_answer("tested") == "tested"

    def test_untested(self):
        assert parse_ge_l4_answer("untested") == "untested"

    def test_with_evidence(self):
        assert parse_ge_l4_answer("tested\nThis gene is in DepMap 22Q2") == "tested"

    def test_case_insensitive(self):
        assert parse_ge_l4_answer("UNTESTED") == "untested"

    def test_embedded(self):
        assert parse_ge_l4_answer("I believe this is untested") == "untested"

    def test_tested_priority_over_untested(self):
        # "untested" contains "tested" — check correct parsing
        assert parse_ge_l4_answer("untested because...") == "untested"

    def test_empty(self):
        assert parse_ge_l4_answer("") is None


class TestEvaluateL4:
    def test_perfect(self):
        preds = ["tested", "untested", "tested"]
        gold = ["tested", "untested", "tested"]
        result = evaluate_ge_l4(preds, gold)
        assert result["accuracy"] == 1.0
        assert result["mcc"] == 1.0

    def test_all_wrong(self):
        preds = ["untested", "tested"]
        gold = ["tested", "untested"]
        result = evaluate_ge_l4(preds, gold)
        assert result["accuracy"] == 0.0

    def test_distribution(self):
        preds = ["tested", "tested", "untested"]
        gold = ["tested", "tested", "untested"]
        result = evaluate_ge_l4(preds, gold)
        assert result["prediction_distribution"]["tested"] == 2
        assert result["prediction_distribution"]["untested"] == 1


# ── Dispatch ──────────────────────────────────────────────────────────────


class TestDispatch:
    def test_l1_dispatch(self):
        result = compute_all_ge_llm_metrics("ge-l1", ["A", "B"], ["A", "B"])
        assert "accuracy" in result

    def test_l4_dispatch(self):
        result = compute_all_ge_llm_metrics("ge-l4", ["tested"], ["tested"])
        assert "accuracy" in result

    def test_l2_dispatch(self):
        pred = '{"genes": ["TP53"], "total_genes_mentioned": 1, "screen_type": "CRISPR"}'
        gold = [{"genes": ["TP53"], "total_genes_mentioned": 1, "screen_type": "CRISPR"}]
        result = compute_all_ge_llm_metrics("ge-l2", [pred], gold)
        assert "parse_rate" in result
        assert "schema_compliance" in result
        assert "field_f1" in result
        assert result["parse_rate"] == 1.0

    def test_l2_dispatch_with_full_record(self):
        pred = '{"genes": ["TP53"], "total_genes_mentioned": 1, "screen_type": "CRISPR"}'
        gold_records = [
            {
                "question_id": "GEL2-001",
                "task": "ge-l2",
                "gold_extraction": {
                    "genes": ["TP53"],
                    "total_genes_mentioned": 1,
                    "screen_type": "CRISPR",
                },
                "gold_answer": "non-essential",
            }
        ]
        result = compute_all_ge_llm_metrics("ge-l2", [pred], gold_records)
        assert result["parse_rate"] == 1.0
        assert "field_f1" in result

    def test_l3_dispatch(self):
        result = compute_all_ge_llm_metrics(
            "ge-l3",
            ["This gene is non-essential because it has redundant paralogs."],
            [{}],
        )
        assert "n_parsed" in result
        assert result["n_parsed"] == 0
        assert result["overall_mean"] == 0.0

    def test_invalid_task(self):
        with pytest.raises(ValueError):
            compute_all_ge_llm_metrics("ge-l99", [], [])