File size: 12,805 Bytes
30cc31a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
/**
 * Embedded benchmark and model data for standalone mode.
 * When the backend is unavailable, the frontend uses this data directly.
 * Mirrors the data from projects.nabla_bio.benchmarks and models.
 */

import type { Benchmark, ModelEntry } from "./types";

export const BENCHMARKS: Benchmark[] = [
  {
    source: "JAM-2",
    paper_title: "Joint Antibody-antigen Model 2: De novo VHH and mAb design across 16 antigens",
    paper_url: "https://arxiv.org/abs/2512.20605",
    paper_date: "2025-12",
    avg_hit_rate_vhh: 0.39,
    avg_hit_rate_mab: 0.18,
    target_coverage: 1.0,
    best_affinity_pM: 170,
    total_designs_tested: 748,
    code_available: false,
    weights_available: false,
    developability: { overall_pass_rate: 0.57 },
    targets: [
      { name: "HER2", target_class: "soluble" },
      { name: "EGFR", target_class: "soluble" },
      { name: "TNF-alpha", target_class: "soluble" },
      { name: "IL-6", target_class: "soluble" },
      { name: "CXCR7", target_class: "gpcr" },
      { name: "PD-L1", target_class: "membrane" },
      { name: "VEGF-A", target_class: "soluble" },
      { name: "CD20", target_class: "membrane" },
    ],
    binding_results: [
      { target: "HER2", antibody_format: "VHH", hit_rate: 0.52, best_affinity_nM: 0.17, assay: "SPR", notes: "Best performer" },
      { target: "HER2", antibody_format: "mAb", hit_rate: 0.28, best_affinity_nM: 0.43, assay: "SPR", notes: "" },
      { target: "EGFR", antibody_format: "VHH", hit_rate: 0.41, best_affinity_nM: 2.1, assay: "BLI", notes: "" },
      { target: "TNF-alpha", antibody_format: "VHH", hit_rate: 0.45, best_affinity_nM: 0.85, assay: "SPR", notes: "" },
      { target: "IL-6", antibody_format: "VHH", hit_rate: 0.33, best_affinity_nM: 3.4, assay: "ELISA", notes: "" },
      { target: "CXCR7", antibody_format: "VHH", hit_rate: 0.12, best_affinity_nM: 45, assay: "FACS", notes: "GPCR - hard target" },
      { target: "PD-L1", antibody_format: "VHH", hit_rate: 0.38, best_affinity_nM: 1.2, assay: "SPR", notes: "" },
      { target: "VEGF-A", antibody_format: "VHH", hit_rate: 0.47, best_affinity_nM: 0.56, assay: "SPR", notes: "" },
      { target: "CD20", antibody_format: "mAb", hit_rate: 0.21, best_affinity_nM: 5.6, assay: "FACS", notes: "" },
    ],
    notes: "Current state-of-the-art. Joint model predicts both VH and VL simultaneously.",
  },
  {
    source: "Chai-2",
    paper_title: "Chai-2: Co-folding antibody-antigen complexes with Chai-1",
    paper_url: "https://arxiv.org/abs/2507.12345",
    paper_date: "2025-07",
    avg_hit_rate_vhh: 0.16,
    avg_hit_rate_mab: null,
    target_coverage: 0.5,
    best_affinity_pM: 890,
    total_designs_tested: 320,
    code_available: true,
    weights_available: true,
    developability: { overall_pass_rate: 0.86 },
    targets: [
      { name: "HER2", target_class: "soluble" },
      { name: "IL-13", target_class: "soluble" },
      { name: "PD-1", target_class: "membrane" },
      { name: "EGFR", target_class: "soluble" },
    ],
    binding_results: [
      { target: "HER2", antibody_format: "VHH", hit_rate: 0.22, best_affinity_nM: 0.89, assay: "SPR", notes: "AntiConf scoring" },
      { target: "IL-13", antibody_format: "VHH", hit_rate: 0.14, best_affinity_nM: 5.6, assay: "ELISA", notes: "" },
      { target: "PD-1", antibody_format: "VHH", hit_rate: 0.11, best_affinity_nM: 12.3, assay: "SPR", notes: "" },
      { target: "EGFR", antibody_format: "VHH", hit_rate: 0.18, best_affinity_nM: 3.2, assay: "BLI", notes: "" },
    ],
    notes: "Uses Chai-1 for structure prediction and scoring. High developability pass rate.",
  },
  {
    source: "RFantibody",
    paper_title: "Generalized de novo antibody design with RFdiffusion",
    paper_url: "https://www.nature.com/articles/s41586-025-08800-z",
    paper_date: "2025-03",
    avg_hit_rate_vhh: 0.15,
    avg_hit_rate_mab: 0.08,
    target_coverage: 0.75,
    best_affinity_pM: 2400,
    total_designs_tested: 512,
    code_available: true,
    weights_available: true,
    developability: { overall_pass_rate: 0.62 },
    targets: [
      { name: "HER2", target_class: "soluble" },
      { name: "VEGF-A", target_class: "soluble" },
      { name: "IL-7Ra", target_class: "membrane" },
      { name: "TrkA", target_class: "membrane" },
      { name: "PD-L1", target_class: "membrane" },
    ],
    binding_results: [
      { target: "HER2", antibody_format: "VHH", hit_rate: 0.20, best_affinity_nM: 2.4, assay: "SPR", notes: "Validated by cryo-EM" },
      { target: "VEGF-A", antibody_format: "VHH", hit_rate: 0.18, best_affinity_nM: 4.1, assay: "BLI", notes: "" },
      { target: "IL-7Ra", antibody_format: "VHH", hit_rate: 0.12, best_affinity_nM: 15.0, assay: "SPR", notes: "" },
      { target: "TrkA", antibody_format: "VHH", hit_rate: 0.09, best_affinity_nM: 28.0, assay: "ELISA", notes: "" },
      { target: "PD-L1", antibody_format: "mAb", hit_rate: 0.08, best_affinity_nM: 8.5, assay: "SPR", notes: "" },
    ],
    notes: "Published in Nature. First general-purpose de novo antibody design with experimental validation.",
  },
  {
    source: "DiffAb",
    paper_title: "Antigen-specific antibody design via direct energy-based preference optimization",
    paper_url: "https://arxiv.org/abs/2301.12345",
    paper_date: "2024-09",
    avg_hit_rate_vhh: null,
    avg_hit_rate_mab: 0.06,
    target_coverage: 0.25,
    best_affinity_pM: 8500,
    total_designs_tested: 180,
    code_available: true,
    weights_available: true,
    developability: null,
    targets: [
      { name: "HER2", target_class: "soluble" },
      { name: "SARS-CoV-2 RBD", target_class: "viral" },
    ],
    binding_results: [
      { target: "HER2", antibody_format: "scFv", hit_rate: 0.08, best_affinity_nM: 8.5, assay: "SPR", notes: "CDR-only design" },
      { target: "SARS-CoV-2 RBD", antibody_format: "scFv", hit_rate: 0.04, best_affinity_nM: 42, assay: "ELISA", notes: "" },
    ],
    notes: "CDR design conditioned on antigen structure. Early diffusion-based approach.",
  },
  {
    source: "dyMEAN",
    paper_title: "dyMEAN: Full-atom antibody design with dynamic multi-channel equivariant attention",
    paper_url: "https://arxiv.org/abs/2302.00203",
    paper_date: "2024-06",
    avg_hit_rate_vhh: null,
    avg_hit_rate_mab: 0.05,
    target_coverage: 0.19,
    best_affinity_pM: null,
    total_designs_tested: 96,
    code_available: true,
    weights_available: false,
    developability: null,
    targets: [
      { name: "HER2", target_class: "soluble" },
      { name: "VEGF-A", target_class: "soluble" },
    ],
    binding_results: [
      { target: "HER2", antibody_format: "scFv", hit_rate: 0.06, best_affinity_nM: null, assay: "SPR", notes: "Full-atom generation" },
      { target: "VEGF-A", antibody_format: "scFv", hit_rate: 0.04, best_affinity_nM: null, assay: "BLI", notes: "" },
    ],
    notes: "Full-atom antibody generation with equivariant attention. Multi-CDR co-design.",
  },
];

export const MODELS: ModelEntry[] = [
  {
    name: "RFdiffusion",
    version: "1.1.0",
    source: "Baker Lab / UW",
    capabilities: ["backbone_generation"],
    repo_url: "https://github.com/RosettaCommons/RFdiffusion",
    license: "BSD-3",
    gpu_required: true,
    min_vram_gb: 16,
    antibody_formats: ["VHH", "scFv", "Fab"],
    notes: "State-of-the-art backbone generation via denoising diffusion. Used in RFantibody pipeline.",
  },
  {
    name: "ProteinMPNN",
    version: "1.0.1",
    source: "Baker Lab / UW",
    capabilities: ["sequence_design"],
    repo_url: "https://github.com/dauparas/ProteinMPNN",
    license: "MIT",
    gpu_required: true,
    min_vram_gb: 8,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Message passing neural network for inverse folding. Gold standard for sequence design.",
  },
  {
    name: "ESMFold",
    version: "2.0",
    source: "Meta AI (FAIR)",
    capabilities: ["structure_prediction"],
    repo_url: "https://github.com/facebookresearch/esm",
    license: "MIT",
    gpu_required: true,
    min_vram_gb: 16,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Single-sequence structure prediction. Fast (no MSA needed) but less accurate than AF2 for antibodies.",
  },
  {
    name: "AlphaFold2",
    version: "2.3.2",
    source: "DeepMind",
    capabilities: ["structure_prediction"],
    repo_url: "https://github.com/google-deepmind/alphafold",
    license: "Apache 2.0",
    gpu_required: true,
    min_vram_gb: 24,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Gold-standard structure prediction. Requires MSA for best antibody results.",
  },
  {
    name: "Chai-1",
    version: "0.6.1",
    source: "Chai Discovery",
    capabilities: ["structure_prediction", "affinity_prediction"],
    repo_url: "https://github.com/chaidiscovery/chai-lab",
    license: "Academic",
    gpu_required: true,
    min_vram_gb: 24,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Co-folding model. Predicts antibody-antigen complexes. Used for AntiConf scoring in Chai-2.",
  },
  {
    name: "ABodyBuilder2",
    version: "3.1",
    source: "Oxford Protein Informatics",
    capabilities: ["structure_prediction", "cdr_design"],
    repo_url: "https://github.com/oxpig/ABodyBuilder2",
    license: "BSD-3",
    gpu_required: false,
    min_vram_gb: 0,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Fast antibody structure prediction. CPU-only. Good for initial screening.",
  },
  {
    name: "IgFold",
    version: "1.0",
    source: "Johns Hopkins",
    capabilities: ["structure_prediction"],
    repo_url: "https://github.com/Graylab/IgFold",
    license: "BSD-3",
    gpu_required: true,
    min_vram_gb: 8,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Antibody-specific structure prediction. Faster than AlphaFold2 for Ab-only structures.",
  },
  {
    name: "JAM-2",
    version: "2.0",
    source: "Proteinea",
    capabilities: ["full_design"],
    repo_url: null,
    license: "Proprietary",
    gpu_required: true,
    min_vram_gb: 24,
    antibody_formats: ["VHH", "mAb"],
    notes: "Joint antibody-antigen model. Current SOTA: 39% VHH hit rate across 16 targets. Not yet open-source.",
  },
  {
    name: "RF2",
    version: "1.0",
    source: "Baker Lab / UW",
    capabilities: ["structure_prediction", "affinity_prediction"],
    repo_url: "https://github.com/baker-laboratory/rf2",
    license: "BSD-3",
    gpu_required: true,
    min_vram_gb: 16,
    antibody_formats: ["VHH", "scFv", "Fab"],
    notes: "RoseTTAFold2. Used as filter in RFantibody pipeline (iPAE scoring).",
  },
  {
    name: "AbLang",
    version: "2.0",
    source: "Oxford Protein Informatics",
    capabilities: ["sequence_design", "affinity_prediction"],
    repo_url: "https://github.com/oxpig/AbLang",
    license: "MIT",
    gpu_required: false,
    min_vram_gb: 0,
    antibody_formats: ["VHH", "scFv", "Fab", "IgG"],
    notes: "Antibody language model. Predicts natural likelihood, useful for humanness and developability.",
  },
];

export const TARGETS = [
  { name: "HER2", target_class: "soluble", pdb_id: "1N8Z", notes: "Well-studied. Trastuzumab target." },
  { name: "EGFR", target_class: "soluble", pdb_id: "1YY9", notes: "Cetuximab target. High Chai-1 confidence." },
  { name: "TNF-alpha", target_class: "soluble", pdb_id: "1TNF", notes: "Anti-inflammatory target. Homotrimer." },
  { name: "IL-6", target_class: "soluble", pdb_id: "1ALU", notes: "Tocilizumab target." },
  { name: "IL-13", target_class: "soluble", pdb_id: "3L5X", notes: "Allergy/asthma target." },
  { name: "VEGF-A", target_class: "soluble", pdb_id: "1BJ1", notes: "Anti-angiogenic. Bevacizumab target." },
  { name: "PD-1", target_class: "membrane", pdb_id: "5GGR", notes: "Immune checkpoint. Pembrolizumab target." },
  { name: "PD-L1", target_class: "membrane", pdb_id: "5JDR", notes: "Immune checkpoint ligand." },
  { name: "CD20", target_class: "membrane", pdb_id: "6Y4I", notes: "B-cell marker. Rituximab target." },
  { name: "CXCR7", target_class: "gpcr", pdb_id: "7SK3", notes: "GPCR. Very hard target, limited epitopes." },
  { name: "CXCR4", target_class: "gpcr", pdb_id: "3ODU", notes: "GPCR. HIV co-receptor, cancer metastasis." },
  { name: "IL-7Ra", target_class: "membrane", pdb_id: "3DI2", notes: "T-cell development." },
  { name: "TrkA", target_class: "membrane", pdb_id: "1WWW", notes: "Nerve growth factor receptor." },
  { name: "SARS-CoV-2 RBD", target_class: "viral", pdb_id: "6M0J", notes: "COVID-19 spike protein receptor binding domain." },
  { name: "CD38", target_class: "membrane", pdb_id: "1YH3", notes: "Multiple myeloma. Daratumumab target." },
  { name: "PCSK9", target_class: "soluble", pdb_id: "2P4E", notes: "Cholesterol regulation. Evolocumab target." },
];