jang1563 commited on
Commit
d84bc62
·
verified ·
1 Parent(s): 915dc8b

Add data exports: LLM benchmarks, CT, PPI small files

Browse files
Files changed (38) hide show
  1. data/README.md +123 -0
  2. data/compound_names.parquet +3 -0
  3. data/croissant.json +176 -0
  4. data/ct/negbiodb_ct_m1_balanced.parquet +3 -0
  5. data/ct/negbiodb_ct_m1_degree_matched.parquet +3 -0
  6. data/ct/negbiodb_ct_m1_realistic.parquet +3 -0
  7. data/ct/negbiodb_ct_m1_smiles_only.parquet +3 -0
  8. data/ct/negbiodb_ct_m1_uniform_random.parquet +3 -0
  9. data/ct/negbiodb_ct_m2.parquet +3 -0
  10. data/ct/negbiodb_ct_pairs.parquet +3 -0
  11. data/ct_llm/ct_l1_dataset.jsonl +0 -0
  12. data/ct_llm/ct_l1_metadata.json +22 -0
  13. data/ct_llm/ct_l2_dataset.jsonl +0 -0
  14. data/ct_llm/ct_l2_metadata.json +24 -0
  15. data/ct_llm/ct_l3_dataset.jsonl +0 -0
  16. data/ct_llm/ct_l3_metadata.json +13 -0
  17. data/ct_llm/ct_l4_dataset.jsonl +0 -0
  18. data/ct_llm/ct_l4_metadata.json +19 -0
  19. data/ge_llm/ge-l1_metadata.json +22 -0
  20. data/ge_llm/ge-l2_metadata.json +17 -0
  21. data/ge_llm/ge-l3_metadata.json +15 -0
  22. data/ge_llm/ge-l4_metadata.json +17 -0
  23. data/ge_llm/ge_l1_dataset.jsonl +0 -0
  24. data/ge_llm/ge_l2_dataset.jsonl +0 -0
  25. data/ge_llm/ge_l3_dataset.jsonl +0 -0
  26. data/ge_llm/ge_l4_dataset.jsonl +0 -0
  27. data/llm_benchmarks/l1_mcq.jsonl +0 -0
  28. data/llm_benchmarks/l2_candidates.jsonl +0 -0
  29. data/llm_benchmarks/l3_reasoning_pilot.jsonl +50 -0
  30. data/llm_benchmarks/l4_tested_untested.jsonl +0 -0
  31. data/ppi_llm/ppi_l1_dataset.jsonl +0 -0
  32. data/ppi_llm/ppi_l1_metadata.json +22 -0
  33. data/ppi_llm/ppi_l2_dataset.jsonl +0 -0
  34. data/ppi_llm/ppi_l2_metadata.json +17 -0
  35. data/ppi_llm/ppi_l3_dataset.jsonl +0 -0
  36. data/ppi_llm/ppi_l3_metadata.json +19 -0
  37. data/ppi_llm/ppi_l4_dataset.jsonl +0 -0
  38. data/ppi_llm/ppi_l4_metadata.json +17 -0
data/README.md ADDED
@@ -0,0 +1,123 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-sa-4.0
3
+ language:
4
+ - en
5
+ tags:
6
+ - drug-target-interaction
7
+ - clinical-trials
8
+ - protein-protein-interaction
9
+ - negative-results
10
+ - benchmark
11
+ - bioinformatics
12
+ task_categories:
13
+ - text-classification
14
+ - question-answering
15
+ - text-generation
16
+ size_categories:
17
+ - 10M<n<100M
18
+ ---
19
+
20
+ # NegBioDB: A Negative Results Database for Biomedical Sciences
21
+
22
+ NegBioDB is a large-scale database of experimentally confirmed negative results across three biomedical domains, paired with dual ML/LLM benchmarks for evaluating how well computational methods handle negative evidence.
23
+
24
+ ## Overview
25
+
26
+ | Domain | Negative Results | Entities | Benchmark Tasks |
27
+ |--------|-----------------|----------|-----------------|
28
+ | DTI (Drug-Target Interaction) | 24,965,618 pairs | 919K compounds, 3.7K targets | ML (M1) + LLM (L1-L4) |
29
+ | CT (Clinical Trial Failure) | 102,850 results | 216K trials, 176K interventions | ML (M1-M2) + LLM (L1-L4) |
30
+ | PPI (Protein-Protein Interaction) | 2,220,786 pairs | 18.4K proteins | ML (M1) |
31
+
32
+ ## Data Sources
33
+
34
+ ### DTI Domain
35
+ | Source | License | Contribution |
36
+ |--------|---------|-------------|
37
+ | ChEMBL 34 | CC BY-SA 3.0 | Curated bioactivity data |
38
+ | PubChem BioAssay | Public domain | HTS screening results |
39
+ | BindingDB | CC BY-SA 3.0 | Binding measurements |
40
+ | DAVIS | CC BY 4.0 | Kinase selectivity panel |
41
+
42
+ ### CT Domain
43
+ | Source | License | Contribution |
44
+ |--------|---------|-------------|
45
+ | AACT (ClinicalTrials.gov) | Public domain | Trial metadata |
46
+ | Open Targets | Apache 2.0 | Drug-target mappings |
47
+ | CTO | MIT | Clinical trial outcomes |
48
+ | Shi & Du 2024 | CC BY 4.0 | Safety/efficacy data |
49
+
50
+ ### PPI Domain
51
+ | Source | License | Contribution |
52
+ |--------|---------|-------------|
53
+ | IntAct | CC BY 4.0 | Curated non-interactions |
54
+ | HuRI | CC BY 4.0 | Y2H systematic negatives |
55
+ | hu.MAP 3.0 | CC BY 4.0 | Complex-derived negatives |
56
+ | STRING v12.0 | CC BY 4.0 | Zero-score protein pairs |
57
+
58
+ ## File Structure
59
+
60
+ ### DTI Files
61
+ | File | Size | Rows | Description |
62
+ |------|------|------|-------------|
63
+ | `negbiodb_dti_pairs.parquet` | 139.4 MB | 24,965,618 | All negative DTI pairs with metadata |
64
+ | `negbiodb_m1_balanced.parquet` | 270.2 MB | 1,725,446 | M1 balanced dataset (1:1 pos:neg) |
65
+ | `negbiodb_m1_realistic.parquet` | 753.3 MB | 9,489,953 | M1 realistic dataset (1:10 pos:neg) |
66
+ | `negbiodb_m1_uniform_random.parquet` | 467.3 MB | 1,767,380 | Control: uniform random negatives |
67
+ | `negbiodb_m1_degree_matched.parquet` | 274.9 MB | 1,767,380 | Control: degree-matched negatives |
68
+ | `negbiodb_m1_balanced_ddb.parquet` | 1.0 GB | 1,725,446 | M1 balanced with degree-balanced split |
69
+
70
+ ### CT Files
71
+ | File | Size | Rows | Description |
72
+ |------|------|------|-------------|
73
+ | `ct/negbiodb_ct_pairs.parquet` | 1.7 MB | 102,850 | All CT failure pairs |
74
+ | `ct/negbiodb_ct_m1_balanced.parquet` | 308.7 KB | 11,222 | CT-M1 balanced (success vs failure) |
75
+ | `ct/negbiodb_ct_m1_realistic.parquet` | 896.4 KB | 36,957 | CT-M1 realistic ratio |
76
+ | `ct/negbiodb_ct_m2.parquet` | 2.2 MB | 112,298 | CT-M2 multiclass failure category |
77
+
78
+ ### PPI Files
79
+ | File | Size | Rows | Description |
80
+ |------|------|------|-------------|
81
+ | `ppi/negbiodb_ppi_pairs.parquet` | 685.2 MB | 2,220,786 | All negative PPI pairs |
82
+ | `ppi/ppi_m1_balanced.parquet` | 120.5 MB | 123,456 | PPI-M1 balanced (1:1 pos:neg) |
83
+ | `ppi/ppi_m1_realistic.parquet` | 744.8 MB | 679,008 | PPI-M1 realistic (1:10 ratio) |
84
+
85
+ ### LLM Benchmark Files (DTI)
86
+ | File | Items | Description |
87
+ |------|-------|-------------|
88
+ | `llm_benchmarks/l1_mcq.jsonl` | 1,942 | L1: 4-class activity MCQ |
89
+ | `llm_benchmarks/l3_reasoning_pilot.jsonl` | 50 | L3: Scientific reasoning (pilot) |
90
+ | `llm_benchmarks/l4_tested_untested.jsonl` | 500 | L4: Tested vs untested discrimination |
91
+
92
+ ## Benchmark Tasks
93
+
94
+ ### ML Benchmarks
95
+ | Task | Domain | Type | Splits |
96
+ |------|--------|------|--------|
97
+ | M1 | DTI | Binary (active/inactive) | random, cold_compound, cold_target, degree_balanced |
98
+ | CT-M1 | CT | Binary (success/failure) | random, cold_drug, cold_condition, temporal, scaffold, cold_both |
99
+ | CT-M2 | CT | 7-way failure category | Same as CT-M1 |
100
+ | PPI-M1 | PPI | Binary (interact/non-interact) | random, cold_protein, cold_both, degree_balanced |
101
+
102
+ ### LLM Benchmarks (DTI)
103
+ | Task | Type | Size | Description |
104
+ |------|------|------|-------------|
105
+ | L1 | MCQ classification | 1,600 | 4-class activity level |
106
+ | L2 | Structured extraction | ~100 | Extract results from abstracts |
107
+ | L3 | Reasoning | 50 | Explain compound-target inactivity |
108
+ | L4 | Discrimination | 400 | Tested vs untested pair |
109
+
110
+ ## License
111
+
112
+ This dataset is released under **CC BY-SA 4.0**, due to the viral clause of ChEMBL's CC BY-SA 3.0 license. See the LICENSE file for details.
113
+
114
+ ## Citation
115
+
116
+ ```bibtex
117
+ @dataset{negbiodb2026,
118
+ title={NegBioDB: A Negative Results Database for Biomedical Sciences},
119
+ author={Jang, Jungwon},
120
+ year={2026},
121
+ url={https://github.com/jang1563/NegBioDB}
122
+ }
123
+ ```
data/compound_names.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d63e26b61ca7d3694ae20c9553d8460c827170557d114253d0e89b71fd5636f9
3
+ size 12400351
data/croissant.json ADDED
@@ -0,0 +1,176 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "@context": {
3
+ "@vocab": "https://schema.org/",
4
+ "sc": "https://schema.org/",
5
+ "cr": "http://mlcommons.org/croissant/",
6
+ "rai": "http://mlcommons.org/croissant/RAI/"
7
+ },
8
+ "@type": "sc:Dataset",
9
+ "name": "NegBioDB",
10
+ "description": "A large-scale database of experimentally confirmed negative results across three biomedical domains (DTI, Clinical Trials, PPI), with dual ML/LLM benchmarks.",
11
+ "license": "https://creativecommons.org/licenses/by-sa/4.0/",
12
+ "url": "https://github.com/jang1563/NegBioDB",
13
+ "version": "1.0.0",
14
+ "datePublished": "2026",
15
+ "creator": {
16
+ "@type": "sc:Person",
17
+ "name": "Jungwon Jang"
18
+ },
19
+ "distribution": [
20
+ {
21
+ "@type": "cr:FileObject",
22
+ "name": "dti_pairs",
23
+ "contentUrl": "negbiodb_dti_pairs.parquet",
24
+ "encodingFormat": "application/x-parquet",
25
+ "description": "All negative DTI pairs with source, tier, and activity data"
26
+ },
27
+ {
28
+ "@type": "cr:FileObject",
29
+ "name": "dti_m1_balanced",
30
+ "contentUrl": "negbiodb_m1_balanced.parquet",
31
+ "encodingFormat": "application/x-parquet",
32
+ "description": "DTI M1 balanced benchmark dataset (1:1 positive:negative)"
33
+ },
34
+ {
35
+ "@type": "cr:FileObject",
36
+ "name": "ct_pairs",
37
+ "contentUrl": "ct/negbiodb_ct_pairs.parquet",
38
+ "encodingFormat": "application/x-parquet",
39
+ "description": "All clinical trial failure pairs"
40
+ },
41
+ {
42
+ "@type": "cr:FileObject",
43
+ "name": "ppi_pairs",
44
+ "contentUrl": "ppi/negbiodb_ppi_pairs.parquet",
45
+ "encodingFormat": "application/x-parquet",
46
+ "description": "All negative PPI pairs"
47
+ },
48
+ {
49
+ "@type": "cr:FileObject",
50
+ "name": "llm_l1",
51
+ "contentUrl": "llm_benchmarks/l1_mcq.jsonl",
52
+ "encodingFormat": "application/jsonl",
53
+ "description": "L1 MCQ classification benchmark"
54
+ },
55
+ {
56
+ "@type": "cr:FileObject",
57
+ "name": "llm_l4",
58
+ "contentUrl": "llm_benchmarks/l4_tested_untested.jsonl",
59
+ "encodingFormat": "application/jsonl",
60
+ "description": "L4 tested/untested discrimination benchmark"
61
+ }
62
+ ],
63
+ "recordSet": [
64
+ {
65
+ "@type": "cr:RecordSet",
66
+ "name": "dti_pairs_record",
67
+ "source": "dti_pairs",
68
+ "field": [
69
+ {
70
+ "@type": "cr:Field",
71
+ "name": "inchikey_connectivity",
72
+ "dataType": "sc:Text",
73
+ "description": "inchikey connectivity"
74
+ },
75
+ {
76
+ "@type": "cr:Field",
77
+ "name": "uniprot_id",
78
+ "dataType": "sc:Text",
79
+ "description": "uniprot id"
80
+ },
81
+ {
82
+ "@type": "cr:Field",
83
+ "name": "activity_type",
84
+ "dataType": "sc:Text",
85
+ "description": "activity type"
86
+ },
87
+ {
88
+ "@type": "cr:Field",
89
+ "name": "pchembl_value",
90
+ "dataType": "sc:Float",
91
+ "description": "pchembl value"
92
+ },
93
+ {
94
+ "@type": "cr:Field",
95
+ "name": "source",
96
+ "dataType": "sc:Text",
97
+ "description": "source"
98
+ },
99
+ {
100
+ "@type": "cr:Field",
101
+ "name": "tier",
102
+ "dataType": "sc:Text",
103
+ "description": "tier"
104
+ }
105
+ ]
106
+ },
107
+ {
108
+ "@type": "cr:RecordSet",
109
+ "name": "ct_pairs_record",
110
+ "source": "ct_pairs",
111
+ "field": [
112
+ {
113
+ "@type": "cr:Field",
114
+ "name": "nct_id",
115
+ "dataType": "sc:Text",
116
+ "description": "nct id"
117
+ },
118
+ {
119
+ "@type": "cr:Field",
120
+ "name": "intervention_name",
121
+ "dataType": "sc:Text",
122
+ "description": "intervention name"
123
+ },
124
+ {
125
+ "@type": "cr:Field",
126
+ "name": "failure_category",
127
+ "dataType": "sc:Text",
128
+ "description": "failure category"
129
+ },
130
+ {
131
+ "@type": "cr:Field",
132
+ "name": "tier",
133
+ "dataType": "sc:Text",
134
+ "description": "tier"
135
+ },
136
+ {
137
+ "@type": "cr:Field",
138
+ "name": "highest_phase_reached",
139
+ "dataType": "sc:Text",
140
+ "description": "highest phase reached"
141
+ }
142
+ ]
143
+ },
144
+ {
145
+ "@type": "cr:RecordSet",
146
+ "name": "ppi_pairs_record",
147
+ "source": "ppi_pairs",
148
+ "field": [
149
+ {
150
+ "@type": "cr:Field",
151
+ "name": "protein_a",
152
+ "dataType": "sc:Text",
153
+ "description": "protein a"
154
+ },
155
+ {
156
+ "@type": "cr:Field",
157
+ "name": "protein_b",
158
+ "dataType": "sc:Text",
159
+ "description": "protein b"
160
+ },
161
+ {
162
+ "@type": "cr:Field",
163
+ "name": "source",
164
+ "dataType": "sc:Text",
165
+ "description": "source"
166
+ },
167
+ {
168
+ "@type": "cr:Field",
169
+ "name": "tier",
170
+ "dataType": "sc:Text",
171
+ "description": "tier"
172
+ }
173
+ ]
174
+ }
175
+ ]
176
+ }
data/ct/negbiodb_ct_m1_balanced.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d948f6ce736bd40ba8d0a65ae741ec3220aa26e29654d49487c0ffb257a97c39
3
+ size 316145
data/ct/negbiodb_ct_m1_degree_matched.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:62d4e4469638dc4c7c45fd5e8992a3b8a8c45efe407b5790396622904ca827b0
3
+ size 428340
data/ct/negbiodb_ct_m1_realistic.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8fb6e691ec5dd69ad8ccbc4447f34981350a8bfb4f811c361b93e9d8cb0bcf7c
3
+ size 917905
data/ct/negbiodb_ct_m1_smiles_only.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95032b1277b1bb3f8740a568b4e4074e1e91b0aca815372721561d83a4ba1ece
3
+ size 174375
data/ct/negbiodb_ct_m1_uniform_random.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5136d02939f49b3ad2725a33924d4b83645c7aefdc68837983bd52dec4bcac3b
3
+ size 483222
data/ct/negbiodb_ct_m2.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7035bf7cf4aecd8b224e7bd8f888dd1c8fd4e5b9ce8d44c7aacb59d189e2a7aa
3
+ size 2313878
data/ct/negbiodb_ct_pairs.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3334c220fa96232e83437683eec7e441e5fca3b313161883aa363217bf99eb88
3
+ size 1752110
data/ct_llm/ct_l1_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ct_llm/ct_l1_metadata.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ct-l1",
3
+ "created": "2026-03-18T18:44:59.269332+00:00",
4
+ "total": 1500,
5
+ "classes": {
6
+ "A": 300,
7
+ "B": 300,
8
+ "C": 300,
9
+ "D": 300,
10
+ "E": 300
11
+ },
12
+ "difficulty": {
13
+ "hard": 856,
14
+ "medium": 524,
15
+ "easy": 120
16
+ },
17
+ "splits": {
18
+ "fewshot": 300,
19
+ "val": 300,
20
+ "test": 900
21
+ }
22
+ }
data/ct_llm/ct_l2_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ct_llm/ct_l2_metadata.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ct-l2",
3
+ "created": "2026-03-18T18:43:54.190316+00:00",
4
+ "total": 500,
5
+ "categories": {
6
+ "other": 69,
7
+ "enrollment": 185,
8
+ "safety": 36,
9
+ "efficacy": 90,
10
+ "regulatory": 10,
11
+ "strategic": 84,
12
+ "design": 26
13
+ },
14
+ "difficulty": {
15
+ "hard": 35,
16
+ "easy": 115,
17
+ "medium": 350
18
+ },
19
+ "splits": {
20
+ "fewshot": 50,
21
+ "val": 50,
22
+ "test": 400
23
+ }
24
+ }
data/ct_llm/ct_l3_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ct_llm/ct_l3_metadata.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ct-l3",
3
+ "created": "2026-03-18T18:43:01.086284+00:00",
4
+ "total": 200,
5
+ "categories": {
6
+ "efficacy": 200
7
+ },
8
+ "splits": {
9
+ "fewshot": 20,
10
+ "val": 20,
11
+ "test": 160
12
+ }
13
+ }
data/ct_llm/ct_l4_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ct_llm/ct_l4_metadata.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ct-l4",
3
+ "created": "2026-03-18T18:46:58.238502+00:00",
4
+ "total": 500,
5
+ "classes": {
6
+ "tested": 250,
7
+ "untested": 250
8
+ },
9
+ "temporal": {
10
+ "post_2023": 125,
11
+ "pre_2020": 125
12
+ },
13
+ "splits": {
14
+ "fewshot": 50,
15
+ "val": 50,
16
+ "test": 400
17
+ },
18
+ "n_leaks": 0
19
+ }
data/ge_llm/ge-l1_metadata.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ge-l1",
3
+ "domain": "ge",
4
+ "n_total": 1200,
5
+ "n_per_class": {
6
+ "common_essential": 300,
7
+ "selective_essential": 300,
8
+ "non_essential": 300,
9
+ "unknown_untested": 300
10
+ },
11
+ "difficulty_distribution": {
12
+ "easy": "480",
13
+ "medium": "420",
14
+ "hard": "300"
15
+ },
16
+ "split_distribution": {
17
+ "test": "720",
18
+ "fewshot": "240",
19
+ "val": "240"
20
+ },
21
+ "seed": 42
22
+ }
data/ge_llm/ge-l2_metadata.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ge-l2",
3
+ "domain": "ge",
4
+ "n_total": 500,
5
+ "design": "constructed_evidence",
6
+ "gene_distribution": {
7
+ "1": "200",
8
+ "2": "200",
9
+ "3": "100"
10
+ },
11
+ "split_distribution": {
12
+ "test": "400",
13
+ "fewshot": "50",
14
+ "val": "50"
15
+ },
16
+ "seed": 42
17
+ }
data/ge_llm/ge-l3_metadata.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ge-l3",
3
+ "domain": "ge",
4
+ "n_total": 200,
5
+ "ne_type_distribution": {
6
+ "reference": 100,
7
+ "context": 100
8
+ },
9
+ "split_distribution": {
10
+ "test": "160",
11
+ "val": "20",
12
+ "fewshot": "20"
13
+ },
14
+ "seed": 42
15
+ }
data/ge_llm/ge-l4_metadata.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ge-l4",
3
+ "domain": "ge",
4
+ "n_total": 475,
5
+ "n_tested": 250,
6
+ "n_untested": 225,
7
+ "n_tested_old": 125,
8
+ "n_tested_new": 125,
9
+ "n_untested_trick": 125,
10
+ "n_untested_obvious": 100,
11
+ "split_distribution": {
12
+ "fewshot": 50,
13
+ "val": 50,
14
+ "test": 375
15
+ },
16
+ "seed": 42
17
+ }
data/ge_llm/ge_l1_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ge_llm/ge_l2_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ge_llm/ge_l3_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ge_llm/ge_l4_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/llm_benchmarks/l1_mcq.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/llm_benchmarks/l2_candidates.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/llm_benchmarks/l3_reasoning_pilot.jsonl ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"class": "reasoning", "compound_name": "(S)-N-(1-((5-(N-(tert- butyl)sulfamoyl)naphthalen-1- yl)amino)-1-oxo-3-phenylpropan- 2-yl)pyrimidine-2-carboxamide::US10968172, ID # 4o", "compound_smiles": "CC(C)(C)NS(=O)(=O)c1cccc2c(NC(=O)[C@H](Cc3ccccc3)NC(=O)c3ncccn3)cccc12", "compound_inchikey": "APIMXUDTMIUDED-QHCPKHFHSA-N", "target_uniprot": "Q70CQ3", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: (S)-N-(1-((5-(N-(tert- butyl)sulfamoyl)naphthalen-1- yl)amino)-1-oxo-3-phenylpropan- 2-yl)pyrimidine-2-carboxamide::US10968172, ID # 4o\nSMILES: CC(C)(C)NS(=O)(=O)c1cccc2c(NC(=O)[C@H](Cc3ccccc3)NC(=O)c3ncccn3)cccc12\nTarget: Q70CQ3, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "fewshot", "question_id": "L3-0000"}
2
+ {"class": "reasoning", "compound_name": "US9409892, 33", "compound_smiles": "N#Cc1nc(C(=O)NC2CCC[C@H]2C(=O)O)c(O)c2ccc(Oc3ccccc3)cc12", "compound_inchikey": "TYSUUCOXMBQYMY-PYUWXLGESA-N", "target_uniprot": "Q96KS0", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: US9409892, 33\nSMILES: N#Cc1nc(C(=O)NC2CCC[C@H]2C(=O)O)c(O)c2ccc(Oc3ccccc3)cc12\nTarget: Q96KS0, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "fewshot", "question_id": "L3-0001"}
3
+ {"class": "reasoning", "compound_name": "US9434724, 24", "compound_smiles": "O=C(OCC12CCN(CC1)CC2)c1c[nH]c2cccc(Br)c12", "compound_inchikey": "MRUNYBDRMJNCIJ-UHFFFAOYSA-N", "target_uniprot": "P11712", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: US9434724, 24\nSMILES: O=C(OCC12CCN(CC1)CC2)c1c[nH]c2cccc(Br)c12\nTarget: P11712, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "fewshot", "question_id": "L3-0002"}
4
+ {"class": "reasoning", "compound_name": "3-methyl-6-(piperidin-1-yl)-2H-naphtho[1,2,3-de]quinoline-2,7(3H)-dione", "compound_smiles": "Cn1c(=O)cc2c3c(c(N4CCCCC4)ccc31)C(=O)c1ccccc1-2", "compound_inchikey": "ZMXVWFFEPNMQNW-UHFFFAOYSA-N", "target_uniprot": "P02675", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: 3-methyl-6-(piperidin-1-yl)-2H-naphtho[1,2,3-de]quinoline-2,7(3H)-dione\nSMILES: Cn1c(=O)cc2c3c(c(N4CCCCC4)ccc31)C(=O)c1ccccc1-2\nTarget: P02675, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "fewshot", "question_id": "L3-0003"}
5
+ {"class": "reasoning", "compound_name": "US9120812, 155", "compound_smiles": "CCN1CC(C)(C)Oc2nc(N3CC4CCC(C3)O4)nc(-c3ccc(NC(=O)Nc4cc(OC)c(F)cc4F)cc3)c2C1=O", "compound_inchikey": "GWMPVSYXOSFCEQ-UHFFFAOYSA-N", "target_uniprot": "P27986", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: US9120812, 155\nSMILES: CCN1CC(C)(C)Oc2nc(N3CC4CCC(C3)O4)nc(-c3ccc(NC(=O)Nc4cc(OC)c(F)cc4F)cc3)c2C1=O\nTarget: P27986, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "fewshot", "question_id": "L3-0004"}
6
+ {"class": "reasoning", "compound_name": "DINACICLIB", "compound_smiles": "CCc1cnn2c(NCc3ccc[n+]([O-])c3)cc(N3CCCC[C@H]3CCO)nc12", "compound_inchikey": "PIMQWRZWLQKKBJ-SFHVURJKSA-N", "target_uniprot": "P36507", "target_gene": "MEK2", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: DINACICLIB\nSMILES: CCc1cnn2c(NCc3ccc[n+]([O-])c3)cc(N3CCCC[C@H]3CCO)nc12\nTarget: MEK2 (P36507), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "val", "question_id": "L3-0005"}
7
+ {"class": "reasoning", "compound_name": "US9149492, 16", "compound_smiles": "CCCCCCCCCCNc1cc(-c2ccccc2)nn1-c1ccccc1", "compound_inchikey": "CJKKEVCKJUHCKP-UHFFFAOYSA-N", "target_uniprot": "O75908", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: US9149492, 16\nSMILES: CCCCCCCCCCNc1cc(-c2ccccc2)nn1-c1ccccc1\nTarget: O75908, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "val", "question_id": "L3-0006"}
8
+ {"class": "reasoning", "compound_name": "3-phenyl-2H-benzo[4,5]thiazolo[3,2-a][1,3,5]triazine-2,4(3H)-dione", "compound_smiles": "O=c1nc2sc3ccccc3n2c(=O)n1-c1ccccc1", "compound_inchikey": "FCPXRIBIDATENX-UHFFFAOYSA-N", "target_uniprot": "O76083", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: 3-phenyl-2H-benzo[4,5]thiazolo[3,2-a][1,3,5]triazine-2,4(3H)-dione\nSMILES: O=c1nc2sc3ccccc3n2c(=O)n1-c1ccccc1\nTarget: O76083, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "val", "question_id": "L3-0007"}
9
+ {"class": "reasoning", "compound_name": "HYDROXYFASUDIL", "compound_smiles": "O=S(=O)(c1cccc2c(O)nccc12)N1CCCNCC1", "compound_inchikey": "ZAVGJDAFCZAWSZ-UHFFFAOYSA-N", "target_uniprot": "Q14683", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: HYDROXYFASUDIL\nSMILES: O=S(=O)(c1cccc2c(O)nccc12)N1CCCNCC1\nTarget: Q14683, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "val", "question_id": "L3-0008"}
10
+ {"class": "reasoning", "compound_name": "TERFENADINE", "compound_smiles": "CC(C)(C)c1ccc(C(O)CCCN2CCC(C(O)(c3ccccc3)c3ccccc3)CC2)cc1", "compound_inchikey": "GUGOEEXESWIERI-UHFFFAOYSA-N", "target_uniprot": "Q03431", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 1, "evidence_quality": "silver", "context_text": "Compound: TERFENADINE\nSMILES: CC(C)(C)c1ccc(C(O)CCCN2CCC(C(O)(c3ccccc3)c3ccccc3)CC2)cc1\nTarget: Q03431, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "val", "question_id": "L3-0009"}
11
+ {"class": "reasoning", "compound_name": "ZARDAVERINE", "compound_smiles": "COc1cc(-c2ccc(=O)[nH]n2)ccc1OC(F)F", "compound_inchikey": "HJMQDJPMQIHLPB-UHFFFAOYSA-N", "target_uniprot": "Q9HCR9", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: ZARDAVERINE\nSMILES: COc1cc(-c2ccc(=O)[nH]n2)ccc1OC(F)F\nTarget: Q9HCR9, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0010"}
12
+ {"class": "reasoning", "compound_name": "3-chloro-N-{2-[6-(trifluoromethyl)pyridin-2-yl]-7H-pyrrolo[2,3-d]pyrimidin-4-yl}pyridin-4-amine::US10399987, Example 40", "compound_smiles": "FC(F)(F)c1cccc(-c2nc(Nc3ccncc3Cl)c3cc[nH]c3n2)n1", "compound_inchikey": "WDJWKQYJNPIICW-UHFFFAOYSA-N", "target_uniprot": "P37173", "target_gene": "TGFBR2", "target_family": "kinase", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: 3-chloro-N-{2-[6-(trifluoromethyl)pyridin-2-yl]-7H-pyrrolo[2,3-d]pyrimidin-4-yl}pyridin-4-amine::US10399987, Example 40\nSMILES: FC(F)(F)c1cccc(-c2nc(Nc3ccncc3Cl)c3cc[nH]c3n2)n1\nTarget: TGFBR2 (P37173), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0011"}
13
+ {"class": "reasoning", "compound_name": "4-(4-Methyl-pyridin-2-ylamino)-piperidine-1-carboxylic acid ethyl ester", "compound_smiles": "CCOC(=O)N1CCC(Nc2cc(C)ccn2)CC1", "compound_inchikey": "LNRMJBWADUSJTA-UHFFFAOYSA-N", "target_uniprot": "P29475", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: 4-(4-Methyl-pyridin-2-ylamino)-piperidine-1-carboxylic acid ethyl ester\nSMILES: CCOC(=O)N1CCC(Nc2cc(C)ccn2)CC1\nTarget: P29475, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0012"}
14
+ {"class": "reasoning", "compound_name": "PF-04217903", "compound_smiles": "OCCn1cc(-c2cnc3nnn(Cc4ccc5ncccc5c4)c3n2)cn1", "compound_inchikey": "PDMUGYOXRHVNMO-UHFFFAOYSA-N", "target_uniprot": "O15197", "target_gene": "EPHB6", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: PF-04217903\nSMILES: OCCn1cc(-c2cnc3nnn(Cc4ccc5ncccc5c4)c3n2)cn1\nTarget: EPHB6 (O15197), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0013"}
15
+ {"class": "reasoning", "compound_name": "BARICITINIB", "compound_smiles": "CCS(=O)(=O)N1CC(CC#N)(n2cc(-c3ncnc4[nH]ccc34)cn2)C1", "compound_inchikey": "XUZMWHLSFXCVMG-UHFFFAOYSA-N", "target_uniprot": "P00519", "target_gene": "ABL1", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: BARICITINIB\nSMILES: CCS(=O)(=O)N1CC(CC#N)(n2cc(-c3ncnc4[nH]ccc34)cn2)C1\nTarget: ABL1 (P00519), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0014"}
16
+ {"class": "reasoning", "compound_name": "US9163008, 71", "compound_smiles": "C[C@H](Nc1cc(C(N)=O)nc(N2CCN(C(c3ccc(F)cc3)c3ccc(F)cc3)CC2)n1)C(N)=O", "compound_inchikey": "SSLDCNYHPUYQRB-HNNXBMFYSA-N", "target_uniprot": "Q15858", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: US9163008, 71\nSMILES: C[C@H](Nc1cc(C(N)=O)nc(N2CCN(C(c3ccc(F)cc3)c3ccc(F)cc3)CC2)n1)C(N)=O\nTarget: Q15858, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0015"}
17
+ {"class": "reasoning", "compound_name": "BAFETINIB", "compound_smiles": "Cc1ccc(NC(=O)c2ccc(CN3CC[C@H](N(C)C)C3)c(C(F)(F)F)c2)cc1Nc1nccc(-c2cncnc2)n1", "compound_inchikey": "ZGBAJMQHJDFTQJ-DEOSSOPVSA-N", "target_uniprot": "Q14839", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: BAFETINIB\nSMILES: Cc1ccc(NC(=O)c2ccc(CN3CC[C@H](N(C)C)C3)c(C(F)(F)F)c2)cc1Nc1nccc(-c2cncnc2)n1\nTarget: Q14839, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0016"}
18
+ {"class": "reasoning", "compound_name": "LENVATINIB", "compound_smiles": "COc1cc2nccc(Oc3ccc(NC(=O)NC4CC4)c(Cl)c3)c2cc1C(N)=O", "compound_inchikey": "WOSKHXYHFSIKNG-UHFFFAOYSA-N", "target_uniprot": "P31751", "target_gene": "AKT2", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: LENVATINIB\nSMILES: COc1cc2nccc(Oc3ccc(NC(=O)NC4CC4)c(Cl)c3)c2cc1C(N)=O\nTarget: AKT2 (P31751), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0017"}
19
+ {"class": "reasoning", "compound_name": "2-((4-(5-chloropyrimidin-2-yl)piperazin-1-yl)methyl)-1-methyl-1H-benzo[d]imidazole", "compound_smiles": "Cn1c(CN2CCN(c3ncc(Cl)cn3)CC2)nc2ccccc21", "compound_inchikey": "IXXZGBOSQDVMTC-UHFFFAOYSA-N", "target_uniprot": "Q13255", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 1, "evidence_quality": "silver", "context_text": "Compound: 2-((4-(5-chloropyrimidin-2-yl)piperazin-1-yl)methyl)-1-methyl-1H-benzo[d]imidazole\nSMILES: Cn1c(CN2CCN(c3ncc(Cl)cn3)CC2)nc2ccccc21\nTarget: Q13255, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0018"}
20
+ {"class": "reasoning", "compound_name": "SID17413282", "compound_smiles": "Cc1ccc(NNS(=O)(=O)c2ccc(C)cc2)cc1", "compound_inchikey": "OSDZCCMUWCNEOP-UHFFFAOYSA-N", "target_uniprot": "O14786", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: SID17413282\nSMILES: Cc1ccc(NNS(=O)(=O)c2ccc(C)cc2)cc1\nTarget: O14786, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0019"}
21
+ {"class": "reasoning", "compound_name": "SID853690", "compound_smiles": "CCCCNC(=O)C1CCCN1C(=O)Nc1cccc(OC)c1", "compound_inchikey": "QUPOBSOOZUZGRS-UHFFFAOYSA-N", "target_uniprot": "P04062", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: SID853690\nSMILES: CCCCNC(=O)C1CCCN1C(=O)Nc1cccc(OC)c1\nTarget: P04062, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0020"}
22
+ {"class": "reasoning", "compound_name": "pyrrolidin-1-yl-[(2S,4S)-4-(5-trifluoromethyl-1,3-dihydro-isoindole-2-carbonyl)-pyrrolidin-2-yl]-methanone", "compound_smiles": "O=C([C@@H]1CN[C@H](C(=O)N2CCCC2)C1)N1Cc2ccc(C(F)(F)F)cc2C1", "compound_inchikey": "ZXYMQVABWZXTNI-BBRMVZONSA-N", "target_uniprot": "Q6V1X1", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: pyrrolidin-1-yl-[(2S,4S)-4-(5-trifluoromethyl-1,3-dihydro-isoindole-2-carbonyl)-pyrrolidin-2-yl]-methanone\nSMILES: O=C([C@@H]1CN[C@H](C(=O)N2CCCC2)C1)N1Cc2ccc(C(F)(F)F)cc2C1\nTarget: Q6V1X1, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0021"}
23
+ {"class": "reasoning", "compound_name": "OSIMERTINIB", "compound_smiles": "C=CC(=O)Nc1cc(Nc2nccc(-c3cn(C)c4ccccc34)n2)c(OC)cc1N(C)CCN(C)C", "compound_inchikey": "DUYJMQONPNNFPI-UHFFFAOYSA-N", "target_uniprot": "O14578", "target_gene": "CIT", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: OSIMERTINIB\nSMILES: C=CC(=O)Nc1cc(Nc2nccc(-c3cn(C)c4ccccc34)n2)c(OC)cc1N(C)CCN(C)C\nTarget: CIT (O14578), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0022"}
24
+ {"class": "reasoning", "compound_name": "CEP-32496", "compound_smiles": "COc1cc2ncnc(Oc3cccc(NC(=O)Nc4cc(C(C)(C)C(F)(F)F)on4)c3)c2cc1OC", "compound_inchikey": "DKNUPRMJNUQNHR-UHFFFAOYSA-N", "target_uniprot": "P36896", "target_gene": "ACVR1B", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: CEP-32496\nSMILES: COc1cc2ncnc(Oc3cccc(NC(=O)Nc4cc(C(C)(C)C(F)(F)F)on4)c3)c2cc1OC\nTarget: ACVR1B (P36896), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0023"}
25
+ {"class": "reasoning", "compound_name": "SID7968677", "compound_smiles": "Cc1cc(NC(=O)CSc2ncnc3ccccc23)no1", "compound_inchikey": "KDDOCAGTMUZHJD-UHFFFAOYSA-N", "target_uniprot": "P11142", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: SID7968677\nSMILES: Cc1cc(NC(=O)CSc2ncnc3ccccc23)no1\nTarget: P11142, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0024"}
26
+ {"class": "reasoning", "compound_name": "PACRITINIB", "compound_smiles": "C1=N/C2=N/c3ccc(OCCN4CCCC4)c(c3)COC/C=C/COCc3cccc(c3)C(=C1)N2", "compound_inchikey": "HWXVIOGONBBTBY-ONEGZZNKSA-N", "target_uniprot": "O75460", "target_gene": "ERN1", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: PACRITINIB\nSMILES: C1=N/C2=N/c3ccc(OCCN4CCCC4)c(c3)COC/C=C/COCc3cccc(c3)C(=C1)N2\nTarget: ERN1 (O75460), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0025"}
27
+ {"class": "reasoning", "compound_name": "ACTB-1003", "compound_smiles": "COCc1c(-c2ccc(NC(=O)Nc3cc(C(F)(F)F)ccc3F)c(F)c2)c2c(N)ncnn2c1CN1CCOCC1", "compound_inchikey": "GZPJCJKUZPUFAL-UHFFFAOYSA-N", "target_uniprot": "P78368", "target_gene": "CSNK1G2", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: ACTB-1003\nSMILES: COCc1c(-c2ccc(NC(=O)Nc3cc(C(F)(F)F)ccc3F)c(F)c2)c2c(N)ncnn2c1CN1CCOCC1\nTarget: CSNK1G2 (P78368), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0026"}
28
+ {"class": "reasoning", "compound_name": "VEMURAFENIB", "compound_smiles": "CCCS(=O)(=O)Nc1ccc(F)c(C(=O)c2c[nH]c3ncc(-c4ccc(Cl)cc4)cc23)c1F", "compound_inchikey": "GPXBXXGIAQBQNI-UHFFFAOYSA-N", "target_uniprot": "Q5VT25", "target_gene": "MRCKA", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: VEMURAFENIB\nSMILES: CCCS(=O)(=O)Nc1ccc(F)c(C(=O)c2c[nH]c3ncc(-c4ccc(Cl)cc4)cc23)c1F\nTarget: MRCKA (Q5VT25), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0027"}
29
+ {"class": "reasoning", "compound_name": "BMS-690514", "compound_smiles": "COc1cccc(Nc2ncnn3ccc(CN4CC[C@@H](N)[C@H](O)C4)c23)c1", "compound_inchikey": "CSGQVNMSRKWUSH-IAGOWNOFSA-N", "target_uniprot": "Q709F0", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: BMS-690514\nSMILES: COc1cccc(Nc2ncnn3ccc(CN4CC[C@@H](N)[C@H](O)C4)c23)c1\nTarget: Q709F0, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0028"}
30
+ {"class": "reasoning", "compound_name": "AZD-1480", "compound_smiles": "Cc1cc(Nc2nc(N[C@@H](C)c3ncc(F)cn3)ncc2Cl)[nH]n1", "compound_inchikey": "PDOQBOJDRPLBQU-QMMMGPOBSA-N", "target_uniprot": "Q9NSD9", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: AZD-1480\nSMILES: Cc1cc(Nc2nc(N[C@@H](C)c3ncc(F)cn3)ncc2Cl)[nH]n1\nTarget: Q9NSD9, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0029"}
31
+ {"class": "reasoning", "compound_name": "3-(5-(((1S,2S)-2-(ethylamino) cyclohexyl)amino)- 1-oxoisoindolin-2-yl)piperidine-2,6-dione::US11185537, Compound I-40", "compound_smiles": "CCN[C@H]1CCCC[C@@H]1Nc1ccc2c(c1)CN(C1CCC(=O)NC1=O)C2=O", "compound_inchikey": "DDATULQPTHGNEF-UBFHEZILSA-N", "target_uniprot": "Q13422", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: 3-(5-(((1S,2S)-2-(ethylamino) cyclohexyl)amino)- 1-oxoisoindolin-2-yl)piperidine-2,6-dione::US11185537, Compound I-40\nSMILES: CCN[C@H]1CCCC[C@@H]1Nc1ccc2c(c1)CN(C1CCC(=O)NC1=O)C2=O\nTarget: Q13422, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0030"}
32
+ {"class": "reasoning", "compound_name": "SID26658848", "compound_smiles": "C=CCOc1ccc(/C=C2\\SC(=S)N(CCC(=O)O)C2=O)cc1", "compound_inchikey": "AEZGVSCOIZTTBC-RAXLEYEMSA-N", "target_uniprot": "Q9Y6Q9", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: SID26658848\nSMILES: C=CCOc1ccc(/C=C2\\SC(=S)N(CCC(=O)O)C2=O)cc1\nTarget: Q9Y6Q9, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0031"}
33
+ {"class": "reasoning", "compound_name": "US9981950, Example 13", "compound_smiles": "Clc1ccc(-n2cc(Cn3cnc(C4CC4)n3)nn2)cc1", "compound_inchikey": "FYLLXYYSVQPAKZ-UHFFFAOYSA-N", "target_uniprot": "Q12809", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: US9981950, Example 13\nSMILES: Clc1ccc(-n2cc(Cn3cnc(C4CC4)n3)nn2)cc1\nTarget: Q12809, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0032"}
34
+ {"class": "reasoning", "compound_name": "APITOLISIB", "compound_smiles": "Cc1c(CN2CCN(C(=O)[C@H](C)O)CC2)sc2c(N3CCOCC3)nc(-c3cnc(N)nc3)nc12", "compound_inchikey": "YOVVNQKCSKSHKT-HNNXBMFYSA-N", "target_uniprot": "Q92918", "target_gene": "HPK1", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: APITOLISIB\nSMILES: Cc1c(CN2CCN(C(=O)[C@H](C)O)CC2)sc2c(N3CCOCC3)nc(-c3cnc(N)nc3)nc12\nTarget: HPK1 (Q92918), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0033"}
35
+ {"class": "reasoning", "compound_name": "THIRAM", "compound_smiles": "CN(C)C(=S)SSC(=S)N(C)C", "compound_inchikey": "KUAZQDVKQLNFPE-UHFFFAOYSA-N", "target_uniprot": "Q6W5P4", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 1, "evidence_quality": "silver", "context_text": "Compound: THIRAM\nSMILES: CN(C)C(=S)SSC(=S)N(C)C\nTarget: Q6W5P4, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0034"}
36
+ {"class": "reasoning", "compound_name": "SID22402056", "compound_smiles": "O=C(CCl)NC(=O)NC12CC3CC(CC(C3)C1)C2", "compound_inchikey": "OBLFEICGDWLKQF-UHFFFAOYSA-N", "target_uniprot": "O15492", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: SID22402056\nSMILES: O=C(CCl)NC(=O)NC12CC3CC(CC(C3)C1)C2\nTarget: O15492, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0035"}
37
+ {"class": "reasoning", "compound_name": "COBIMETINIB", "compound_smiles": "O=C(c1ccc(F)c(F)c1Nc1ccc(I)cc1F)N1CC(O)([C@@H]2CCCCN2)C1", "compound_inchikey": "BSMCAPRUBJMWDF-KRWDZBQOSA-N", "target_uniprot": "Q7Z406", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: COBIMETINIB\nSMILES: O=C(c1ccc(F)c(F)c1Nc1ccc(I)cc1F)N1CC(O)([C@@H]2CCCCN2)C1\nTarget: Q7Z406, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0036"}
38
+ {"class": "reasoning", "compound_name": "2-(3-chloro-N-(2-chloroacetyl)-4-methoxy-anilino)-N-(2-phenylethyl)-2-(2-thienyl)acetamide", "compound_smiles": "COc1ccc(N(C(=O)CCl)C(C(=O)NCCc2ccccc2)c2cccs2)cc1Cl", "compound_inchikey": "UNVKYJSNMVDZJE-UHFFFAOYSA-N", "target_uniprot": "P41145", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 1, "evidence_quality": "silver", "context_text": "Compound: 2-(3-chloro-N-(2-chloroacetyl)-4-methoxy-anilino)-N-(2-phenylethyl)-2-(2-thienyl)acetamide\nSMILES: COc1ccc(N(C(=O)CCl)C(C(=O)NCCc2ccccc2)c2cccs2)cc1Cl\nTarget: P41145, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0037"}
39
+ {"class": "reasoning", "compound_name": "US11584714, Compound 1471", "compound_smiles": "C[C@H](NC(=O)[C@H]1C[C@H](c2ccccc2)CN1)C(=O)NCc1cnc2sccc2c1", "compound_inchikey": "QOGGVKMWFMLWTF-ZOCIIQOWSA-N", "target_uniprot": "P00734", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: US11584714, Compound 1471\nSMILES: C[C@H](NC(=O)[C@H]1C[C@H](c2ccccc2)CN1)C(=O)NCc1cnc2sccc2c1\nTarget: P00734, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0038"}
40
+ {"class": "reasoning", "compound_name": "SID24780368", "compound_smiles": "Cc1ccc(C)c(-n2cc(CNCCc3cnccn3)c(-c3ccccc3)n2)c1", "compound_inchikey": "FFAQZBJIGLDKCL-UHFFFAOYSA-N", "target_uniprot": "P41143", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: SID24780368\nSMILES: Cc1ccc(C)c(-n2cc(CNCCc3cnccn3)c(-c3ccccc3)n2)c1\nTarget: P41143, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0039"}
41
+ {"class": "reasoning", "compound_name": "(R)-7-fluoro-6-methoxy-N-(1'-azaspiro[cyclopropane-1,2'-bicyclo[2.2.2]octan]-3'-yl)benzo[b]thiophene-2-carboxamide::US10183938, Compound (R)-155", "compound_smiles": "COc1ccc2cc(C(=O)N[C@@H]3C4CCN(CC4)C34CC4)sc2c1F", "compound_inchikey": "WIHCHZXZKLIHFS-QGZVFWFLSA-N", "target_uniprot": "P46098", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: (R)-7-fluoro-6-methoxy-N-(1'-azaspiro[cyclopropane-1,2'-bicyclo[2.2.2]octan]-3'-yl)benzo[b]thiophene-2-carboxamide::US10183938, Compound (R)-155\nSMILES: COc1ccc2cc(C(=O)N[C@@H]3C4CCN(CC4)C34CC4)sc2c1F\nTarget: P46098, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0040"}
42
+ {"class": "reasoning", "compound_name": "SID26665079", "compound_smiles": "COc1ccc2c3c([nH]c2c1)C1CC2C(CC(O)C(OC)C2CO)CN1CC3", "compound_inchikey": "YCLAQLTVHNAEEQ-UHFFFAOYSA-N", "target_uniprot": "Q9NZJ5", "target_gene": null, "target_family": "protein", "num_assays": 3, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: SID26665079\nSMILES: COc1ccc2c3c([nH]c2c1)C1CC2C(CC(O)C(OC)C2CO)CN1CC3\nTarget: Q9NZJ5, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0041"}
43
+ {"class": "reasoning", "compound_name": "MERESTINIB", "compound_smiles": "Cc1ccc(C(=O)Nc2ccc(Oc3cc4cnn(C)c4cc3-c3cn[nH]c3)c(F)c2)c(=O)n1-c1ccc(F)cc1", "compound_inchikey": "QHADVLVFMKEIIP-UHFFFAOYSA-N", "target_uniprot": "P23443", "target_gene": "S6K1", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: MERESTINIB\nSMILES: Cc1ccc(C(=O)Nc2ccc(Oc3cc4cnn(C)c4cc3-c3cn[nH]c3)c(F)c2)c(=O)n1-c1ccc(F)cc1\nTarget: S6K1 (P23443), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0042"}
44
+ {"class": "reasoning", "compound_name": "SID49645791", "compound_smiles": "Cc1coc2c1c(C)cc1oc(=O)c(CC(=O)NCCCCCC(=O)O)c(C)c12", "compound_inchikey": "CSLQCSIPGCWTGF-UHFFFAOYSA-N", "target_uniprot": "P49862", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 3, "evidence_quality": "silver", "context_text": "Compound: SID49645791\nSMILES: Cc1coc2c1c(C)cc1oc(=O)c(CC(=O)NCCCCCC(=O)O)c(C)c12\nTarget: P49862, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0043"}
45
+ {"class": "reasoning", "compound_name": "SID17412990", "compound_smiles": "CC(=O)NS(=O)(=O)c1ccc(NC(=S)NC(=O)c2ccco2)cc1", "compound_inchikey": "JBEDJYDIYZFRRT-UHFFFAOYSA-N", "target_uniprot": "O75496", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 1, "evidence_quality": "silver", "context_text": "Compound: SID17412990\nSMILES: CC(=O)NS(=O)(=O)c1ccc(NC(=S)NC(=O)c2ccco2)cc1\nTarget: O75496, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0044"}
46
+ {"class": "reasoning", "compound_name": "WHI-P131", "compound_smiles": "COc1cc2ncnc(Nc3ccc(O)cc3)c2cc1OC", "compound_inchikey": "HOZUXBLMYUPGPZ-UHFFFAOYSA-N", "target_uniprot": "Q13043", "target_gene": "MST1", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: WHI-P131\nSMILES: COc1cc2ncnc(Nc3ccc(O)cc3)c2cc1OC\nTarget: MST1 (Q13043), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0045"}
47
+ {"class": "reasoning", "compound_name": "SID17511938", "compound_smiles": "COc1ccc(C2Nc3ccccc3C(=O)N2c2ccc(F)cc2)cc1CSc1ccccn1", "compound_inchikey": "UTKZKGGRJXHSHT-UHFFFAOYSA-N", "target_uniprot": "P47898", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: SID17511938\nSMILES: COc1ccc(C2Nc3ccccc3C(=O)N2c2ccc(F)cc2)cc1CSc1ccccn1\nTarget: P47898, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0046"}
48
+ {"class": "reasoning", "compound_name": "APITOLISIB", "compound_smiles": "Cc1c(CN2CCN(C(=O)[C@H](C)O)CC2)sc2c(N3CCOCC3)nc(-c3cnc(N)nc3)nc12", "compound_inchikey": "YOVVNQKCSKSHKT-HNNXBMFYSA-N", "target_uniprot": "Q9Y572", "target_gene": null, "target_family": "protein", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: APITOLISIB\nSMILES: Cc1c(CN2CCN(C(=O)[C@H](C)O)CC2)sc2c(N3CCOCC3)nc(-c3cnc(N)nc3)nc12\nTarget: Q9Y572, protein\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0047"}
49
+ {"class": "reasoning", "compound_name": "ALECTINIB", "compound_smiles": "CCc1cc2c(cc1N1CCC(N3CCOCC3)CC1)C(C)(C)c1[nH]c3cc(C#N)ccc3c1C2=O", "compound_inchikey": "KDGFLJKFZUIJMX-UHFFFAOYSA-N", "target_uniprot": "Q9Y6R4", "target_gene": "MAP3K4", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: ALECTINIB\nSMILES: CCc1cc2c(cc1N1CCC(N3CCOCC3)CC1)C(C)(C)c1[nH]c3cc(C#N)ccc3c1C2=O\nTarget: MAP3K4 (Q9Y6R4), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0048"}
50
+ {"class": "reasoning", "compound_name": "CAPMATINIB", "compound_smiles": "CNC(=O)c1ccc(-c2cnc3ncc(Cc4ccc5ncccc5c4)n3n2)cc1F", "compound_inchikey": "LIOLIMKSCNQPLV-UHFFFAOYSA-N", "target_uniprot": "P53671", "target_gene": "LIMK2", "target_family": "kinase", "num_assays": 2, "num_sources": 2, "evidence_quality": "silver", "context_text": "Compound: CAPMATINIB\nSMILES: CNC(=O)c1ccc(-c2cnc3ncc(Cc4ccc5ncccc5c4)n3n2)cc1F\nTarget: LIMK2 (P53671), kinase\n\nThis compound has been experimentally confirmed as INACTIVE against this target.\n\nExplain the likely molecular and pharmacological reasons for this inactivity.\nConsider: binding site compatibility, selectivity profile, structural features,\nmechanism of action, and any known SAR (structure-activity relationship) data.", "split": "test", "question_id": "L3-0049"}
data/llm_benchmarks/l4_tested_untested.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ppi_llm/ppi_l1_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ppi_llm/ppi_l1_metadata.json ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ppi-l1",
3
+ "created": "2026-03-21T01:04:59.015966+00:00",
4
+ "n_total": 1200,
5
+ "n_per_class": {
6
+ "direct_experimental": 300,
7
+ "systematic_screen": 300,
8
+ "computational_inference": 300,
9
+ "database_absence": 300
10
+ },
11
+ "difficulty_distribution": {
12
+ "easy": "480",
13
+ "medium": "420",
14
+ "hard": "300"
15
+ },
16
+ "split_distribution": {
17
+ "test": "720",
18
+ "fewshot": "240",
19
+ "val": "240"
20
+ },
21
+ "seed": 42
22
+ }
data/ppi_llm/ppi_l2_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ppi_llm/ppi_l2_metadata.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ppi-l2",
3
+ "created": "2026-03-21T01:08:26.491226+00:00",
4
+ "n_total": 500,
5
+ "design": "constructed_evidence_fallback",
6
+ "pair_distribution": {
7
+ "1": "200",
8
+ "2": "200",
9
+ "3": "100"
10
+ },
11
+ "split_distribution": {
12
+ "test": "400",
13
+ "fewshot": "50",
14
+ "val": "50"
15
+ },
16
+ "seed": 42
17
+ }
data/ppi_llm/ppi_l3_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ppi_llm/ppi_l3_metadata.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ppi-l3",
3
+ "created": "2026-03-22T07:45:41.330745+00:00",
4
+ "n_total": 200,
5
+ "compartment_distribution": {
6
+ "true": "100",
7
+ "false": "100"
8
+ },
9
+ "split_distribution": {
10
+ "test": "160",
11
+ "fewshot": "20",
12
+ "val": "20"
13
+ },
14
+ "source_distribution": {
15
+ "humap": "158",
16
+ "huri": "42"
17
+ },
18
+ "seed": 42
19
+ }
data/ppi_llm/ppi_l4_dataset.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
data/ppi_llm/ppi_l4_metadata.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "task": "ppi-l4",
3
+ "created": "2026-03-21T01:14:59.583565+00:00",
4
+ "n_total": 500,
5
+ "n_tested": 250,
6
+ "n_untested": 250,
7
+ "n_tested_pre_2015": 125,
8
+ "n_tested_post_2020": 125,
9
+ "n_untested_trick": 125,
10
+ "n_untested_obvious": 125,
11
+ "split_distribution": {
12
+ "fewshot": 50,
13
+ "val": 50,
14
+ "test": 400
15
+ },
16
+ "seed": 42
17
+ }