Sravankumarbonthada commited on
Commit
69f4389
·
verified ·
1 Parent(s): f75b085

Replace with audited BioFormBench-Real (49 rows, sourced) and new BioFormBench-Marketed (165 rows, 80 sequenced antibodies); remove dev scaffolding and unsourced placeholder data

Browse files
README.md CHANGED
@@ -1,230 +1,203 @@
1
- ---
2
- dataset_info:
3
- features:
4
- - name: protein_name
5
- dtype: string
6
- description: Protein identifier (e.g., P1_IgG, P2_scFv, P3_Fab)
7
- - name: protein_mw
8
- dtype: float32
9
- description: Molecular weight in kDa
10
- - name: protein_pi
11
- dtype: float32
12
- description: Isoelectric point (pH)
13
- - name: protein_tm_baseline
14
- dtype: float32
15
- description: Baseline melting temperature (°C)
16
- - name: buffer_species
17
- dtype: string
18
- description: Buffer type (histidine, phosphate, acetate, citrate)
19
- - name: buffer_ph
20
- dtype: float32
21
- description: Buffer pH value
22
- - name: ionic_strength_mm
23
- dtype: float32
24
- description: Ionic strength in mM
25
- - name: stabilizer_type
26
- dtype: string
27
- description: Stabilizer chemical name (trehalose, sucrose, sorbitol, glycerol, etc.)
28
- - name: stabilizer_conc_percent
29
- dtype: float32
30
- description: Stabilizer concentration (% w/v)
31
- - name: stability_outcome
32
- dtype: float32
33
- description: Measured stability score (0-1 scale, higher = more stable)
34
- - name: stability_metric
35
- dtype: string
36
- description: Measurement type (DSF_Tm_shift, SEC_aggregation_percent, turbidity)
37
- - name: temperature_C
38
- dtype: float32
39
- description: Storage temperature condition (°C)
40
- - name: timepoint_days
41
- dtype: int32
42
- description: Measurement timepoint (days)
43
- - name: source_paper
44
- dtype: string
45
- description: Literature source citation
46
- splits:
47
- - name: train
48
- num_examples: 49
49
- - name: test_lopo
50
- num_examples: 18
51
- - name: full
52
- num_examples: 67
53
- license: cc-by-4.0
54
- ---
55
-
56
- # BioFormBench: Benchmark for Generative Biologics Formulation Design
57
-
58
- ## Dataset Description
59
-
60
- **BioFormBench** is a curated, open-access benchmark of **67 real biologics formulations** extracted from peer-reviewed literature. It enables rigorous evaluation of generative models for de novo formulation design under extreme data scarcity.
61
-
62
- ### Dataset Statistics
63
-
64
- | Property | Value |
65
- |----------|-------|
66
- | **Total formulations** | 67 |
67
- | **LOPO evaluation set** | 18 (3 proteins × 6 each) |
68
- | **Proteins** | 15+ distinct biologics (IgGs, scFvs, Fabs, etc.) |
69
- | **Buffer species** | 6 types (histidine, phosphate, acetate, citrate, etc.) |
70
- | **Stabilizers** | 8+ types (trehalose, sucrose, sorbitol, glycerol, etc.) |
71
- | **pH range** | 4.5–7.5 |
72
- | **Ionic strength range** | 50–300 mM |
73
- | **Stability metrics** | DSF Tm-shift, SEC % aggregation, turbidity |
74
-
75
- ## Use Cases
76
-
77
- 1. **Train generative formulation design models** on synthetic data, evaluate on real BioFormBench
78
- 2. **Leave-one-protein-out (LOPO) cross-validation** to test few-shot adaptation
79
- 3. **Benchmark predictive baselines** (Random Forest, SVM, MLP) against generative approaches
80
- 4. **Ablation studies** to measure impact of architecture components
81
- 5. **Open-science resource** for formulation design research community
82
-
83
- ## Dataset Construction
84
-
85
- ### Curation Process
86
-
87
- 1. **Literature mining:** Systematic search of peer-reviewed papers on biologics formulation screening
88
- 2. **Extraction:** DSF (differential scanning fluorimetry), SEC (size-exclusion chromatography), turbidity data from supplementary materials
89
- 3. **Validation:** Cross-check protein properties (MW, pI, Tm) against UniProt; discard inconsistent entries
90
- 4. **Standardization:** Normalize all measurements to 0–1 scale (higher = more stable)
91
-
92
- ### Data Quality
93
-
94
- - ✅ **Zero errors:** All 67 entries manually verified
95
- - ✅ **Reproducible:** Linked to original papers via DOI
96
- - ✅ **Balanced:** 3–6 formulations per protein (average 4.5)
97
- - ✅ **Diverse:** Multiple buffer types, pH values, stabilizer choices per protein
98
-
99
- ### Proteins Included (LOPO Evaluation Set)
100
-
101
- | Protein | Type | MW (kDa) | pI | # Formulations | Source |
102
- |---------|------|----------|-----|---|---|
103
- | **P1** | IgG (humanized) | 150 | 6.2 | 6 | Lit. mining |
104
- | **P2** | scFv (single-chain Fv) | 27 | 5.8 | 6 | Lit. mining |
105
- | **P3** | Fab (fragment antigen-binding) | 50 | 5.5 | 6 | Lit. mining |
106
-
107
- ## Data Format
108
-
109
- ### CSV Structure
110
-
111
- ```csv
112
- protein_name,protein_mw,protein_pi,protein_tm_baseline,buffer_species,buffer_ph,ionic_strength_mm,stabilizer_type,stabilizer_conc_percent,stability_outcome,stability_metric,temperature_C,timepoint_days,source_paper
113
-
114
- P1_IgG,150,6.2,70.0,phosphate,6.5,150,trehalose,50,0.92,DSF_Tm_shift,4,365,Arakawa2007_JPS100_1692
115
- P1_IgG,150,6.2,70.0,histidine,6.0,100,sucrose,100,0.88,SEC_aggregation,4,180,Arakawa2007_JPS100_1692
116
- P2_scFv,27,5.8,65.0,acetate,5.5,80,sorbitol,75,0.85,turbidity,25,90,Smith2023_PharmaRes40_2341
117
- ...
118
- ```
119
-
120
- ### Column Descriptions
121
-
122
- - **protein_name:** Identifier combining type and name
123
- - **protein_mw:** Molecular weight in kDa
124
- - **protein_pi:** Isoelectric point (pH at zero net charge)
125
- - **protein_tm_baseline:** Melting temperature without stabilizers
126
- - **buffer_species:** Buffer chemical (histidine, phosphate, acetate, citrate)
127
- - **buffer_ph:** pH value of formulation
128
- - **ionic_strength_mm:** Salt concentration (mM)
129
- - **stabilizer_type:** Osmolyte/excipient (trehalose, sucrose, sorbitol, glycerol, etc.)
130
- - **stabilizer_conc_percent:** Stabilizer concentration (% w/v)
131
- - **stability_outcome:** Normalized score [0-1] where 1 = most stable
132
- - **stability_metric:** Measurement method (DSF, SEC, turbidity, etc.)
133
- - **temperature_C:** Storage temperature (°C)
134
- - **timepoint_days:** Duration of storage (days)
135
- - **source_paper:** Original publication cite key
136
-
137
- ## Splits
138
-
139
- ### Train Split (49 formulations)
140
- - Proteins with 1-3 formulations per protein
141
- - Used for **baseline training** and **model pretraining validation**
142
- - Sparse but real data
143
-
144
- ### Test Split: LOPO (18 formulations)
145
- - Exactly 3 proteins × 6 formulations each
146
- - **Leave-one-protein-out cross-validation:**
147
- - Fold 1: Train on P2+P3 (12 formulations), test on P1 (6 formulations)
148
- - Fold 2: Train on P1+P3 (12 formulations), test on P2 (6 formulations)
149
- - Fold 3: Train on P1+P2 (12 formulations), test on P3 (6 formulations)
150
- - Evaluates **few-shot adaptation** to novel proteins
151
-
152
- ### Full Split (67 formulations)
153
- - All collected formulations
154
- - Used for exploratory analysis, future expansion
155
-
156
- ## Benchmark Results
157
-
158
- Models evaluated on BioFormBench LOPO:
159
-
160
- | Model | Recall@10 | Calibration (r) | Coverage (%) |
161
- |-------|-----------|---|---|
162
- | **BioForm-LM (full)** | 0.167 | 0.267 ± 0.660 | 35.3% |
163
- | Random Forest (baseline) | 0.08 | 0.15 | 12.5% |
164
- | SVM (baseline) | 0.05 | 0.10 | 8.3% |
165
- | Random sampling | — | — | 7.5% |
166
-
167
- ## Ethical Considerations
168
-
169
- - ✅ **Public data only:** All formulations from published, peer-reviewed literature
170
- - ✅ **No proprietary data:** No private pharmaceutical data included
171
- - ✅ **Academic use:** Designed for research, not commercial deployment
172
- - ✅ **Computational only:** No real protein/drug data, only published properties
173
-
174
- Generated formulations are **research hypotheses**, not clinical recommendations.
175
-
176
- ## Citation
177
-
178
- ```bibtex
179
- @dataset{kumar2026bioformbench,
180
- title={BioFormBench: Benchmark for Generative Biologics Formulation Design},
181
- author={Kumar, Bonthada Sravan},
182
- year={2026},
183
- publisher={Hugging Face Datasets},
184
- url={https://huggingface.co/datasets/Sravankumarbonthada/BioFormBench}
185
- }
186
- ```
187
-
188
- ## Download & Access
189
-
190
- ### Via Hugging Face Datasets
191
- ```python
192
- from datasets import load_dataset
193
-
194
- # Load full dataset
195
- dataset = load_dataset("Sravankumarbonthada/BioFormBench")
196
-
197
- # Load specific split
198
- train_data = load_dataset("Sravankumarbonthada/BioFormBench", split="train")
199
- test_data = load_dataset("Sravankumarbonthada/BioFormBench", split="test_lopo")
200
- ```
201
-
202
- ### Via GitHub
203
- ```bash
204
- git clone https://github.com/Maheshbonthada/bioform-lm.git
205
- cd bioform-lm
206
- # See data/bioformbench_extracted_real_*.csv
207
- ```
208
-
209
- ## Future Expansion
210
-
211
- **Planned scalings:**
212
- - Target: 200+ formulations across 8-10 proteins
213
- - Systematic literature mining of supplementary materials
214
- - Inclusion of immunogenicity + manufacturability data
215
- - Temporal stability curves (not just endpoint measurements)
216
-
217
- ## Contact
218
-
219
- **Dataset Curator:** Bonthada Sravan Kumar
220
- Email: sravansaijohn@gmail.com
221
- Questions: Open issue on GitHub (Maheshbonthada/bioform-lm)
222
-
223
- ## License
224
-
225
- **CC-BY-4.0** - Attribution required for academic/commercial use
226
-
227
- ---
228
-
229
- **Last Updated:** September 7, 2026
230
- **Status:** ✅ Released | 🔄 Accepting contributions via GitHub
 
1
+ ---
2
+ dataset_info:
3
+ - config_name: bioformbench_real
4
+ features:
5
+ - name: protein_id
6
+ dtype: string
7
+ description: Protein identifier (traceable to source_id)
8
+ - name: protein_mw_kda
9
+ dtype: float32
10
+ description: Molecular weight (kDa)
11
+ - name: protein_pi
12
+ dtype: float32
13
+ description: Isoelectric point
14
+ - name: protein_tm_baseline_c
15
+ dtype: float32
16
+ description: Baseline melting temperature (C), no stabilizer
17
+ - name: buffer_species
18
+ dtype: string
19
+ description: Buffer chemical
20
+ - name: buffer_conc_mm
21
+ dtype: float32
22
+ description: Buffer concentration (mM)
23
+ - name: ph
24
+ dtype: float32
25
+ description: Formulation pH
26
+ - name: ionic_strength_mm
27
+ dtype: float32
28
+ description: Ionic strength (mM)
29
+ - name: osmolarity_mosm_kg
30
+ dtype: float32
31
+ description: Osmolarity (mOsm/kg)
32
+ - name: stabilizers_json
33
+ dtype: string
34
+ description: JSON dict of stabilizer name to concentration
35
+ - name: temperature_c
36
+ dtype: float32
37
+ description: Storage/measurement temperature (C)
38
+ - name: stability_score
39
+ dtype: float32
40
+ description: Normalized stability outcome, 0-1, higher = more stable
41
+ - name: measured_aggregation_percent
42
+ dtype: float32
43
+ description: SEC percent aggregation, where reported
44
+ - name: measured_tm_shift_c
45
+ dtype: float32
46
+ description: DSF Tm shift versus baseline, where reported
47
+ - name: source_title
48
+ dtype: string
49
+ description: Title of the primary literature source (required; unsourced rows are excluded, see Data Integrity)
50
+ - name: source_id
51
+ dtype: string
52
+ description: PubMed Central identifier of the primary source
53
+ splits:
54
+ - name: full
55
+ num_examples: 49
56
+ - config_name: bioformbench_marketed
57
+ features:
58
+ - name: inn
59
+ dtype: string
60
+ description: Antibody International Nonproprietary Name
61
+ - name: title
62
+ dtype: string
63
+ description: FDA label title (product name, manufacturer)
64
+ - name: ph
65
+ dtype: float32
66
+ description: Formulation pH parsed from the label
67
+ - name: buffer_species
68
+ dtype: string
69
+ description: Buffer chemical, where fully resolved
70
+ - name: buffer_conc_mm
71
+ dtype: float32
72
+ description: Buffer concentration (mM), where fully resolved
73
+ - name: surfactant
74
+ dtype: string
75
+ description: Surfactant identity, where present
76
+ - name: fv_pi
77
+ dtype: float32
78
+ description: Isoelectric point computed from the VH+VL Fv sequence (Biopython ProtParam)
79
+ - name: fv_mw_kda
80
+ dtype: float32
81
+ description: Fv molecular weight computed from sequence
82
+ - name: source_text
83
+ dtype: string
84
+ description: Verbatim label sentence(s) the row was parsed from
85
+ splits:
86
+ - name: full
87
+ num_examples: 165
88
+ license: cc-by-4.0
89
+ ---
90
+
91
+ # BioFormBench: Two Open Benchmarks for Biologics Formulation Research
92
+
93
+ This release contains **two** datasets, kept separate because they answer different
94
+ questions and neither should be read as a substitute for the other.
95
+
96
+ ## Data integrity note (please read before using either dataset)
97
+
98
+ An earlier internal draft of BioFormBench-Real contained 67 rows. Auditing
99
+ `source_title`/`source_id` provenance found that **18 of those rows had no
100
+ citation** and used placeholder identifiers (`protein_1`–`protein_5`) with
101
+ round, unverified descriptors (50.0 kDa / pI 7.2 / Tm 65.0 C) repeated across
102
+ otherwise-different molecules. Those rows have been **removed**. The 49 rows in
103
+ this release all trace to a PubMed Central ID and were checked against their
104
+ source text. If you have a copy of the 67-row file, discard it and use this one.
105
+
106
+ ## 1. BioFormBench-Real (49 rows, 13 proteins)
107
+
108
+ Literature-mined formulation-outcome measurements (DSF Tm-shift, SEC %
109
+ aggregation, turbidity) for 13 real proteins across 11 primary papers, each row
110
+ traceable to its source. This is the dataset to use for any claim about
111
+ *stability-optimal* formulation, but at n=13 proteins (9 with enough
112
+ per-protein data for leave-one-protein-out evaluation) it is underpowered for a
113
+ definitive protein-conditionality verdict — see the companion paper's
114
+ Discussion for what scaling this further would require.
115
+
116
+ ## 2. BioFormBench-Marketed (165 rows, 80 antibodies) — new in this release
117
+
118
+ 165 formulation presentations for 80 FDA-approved antibody therapeutics,
119
+ deterministically parsed from FDA Structured Product Labeling (DailyMed) and
120
+ linked to real VH/VL variable-domain sequences from Thera-SAbDab. Every parsed
121
+ field keeps its verbatim source sentence (`source_text`) for audit. To our
122
+ knowledge this is the first open dataset connecting therapeutic antibody
123
+ sequence to marketed formulation composition at this scale.
124
+
125
+ This dataset encodes **what was chosen** for an approved product, not what is
126
+ necessarily stability-optimal — marketed choices are also shaped by
127
+ manufacturability, prior platform investment and regulatory precedent. Used at
128
+ full sample size (n=80) with proper multiple-comparison correction and
129
+ leave-one-protein-out validation, protein sequence identity does **not**
130
+ detectably predict the chosen pH or buffer beyond a constant "platform"
131
+ baseline (all sign-flip p > 0.8) — see the companion paper for the full
132
+ statistical analysis, including a worked example of how a naive n=27
133
+ subsample of this same data produces a spurious, non-reproducible positive
134
+ result (Spearman rho=-0.464, p=0.015, uncorrected) that a full-sample,
135
+ corrected, out-of-sample analysis overturns.
136
+
137
+ ## Use cases
138
+
139
+ 1. Test protein-conditional formulation hypotheses at adequate statistical
140
+ power, with the selection-artifact and platform-baseline results in the
141
+ companion paper as a required comparison point, not merely a possible one.
142
+ 2. Train/evaluate predictive or generative formulation models against a
143
+ documented platform baseline rather than only random/uniform baselines.
144
+ 3. Audit other mechanistic formulation simulators using the diagnostic
145
+ described in the companion paper (dead-slot spread, interior-argmax
146
+ fraction) against these real formulations.
147
+ 4. Sequence-to-formulation representation learning (BioFormBench-Marketed
148
+ includes Fv-derived descriptors so no additional sequence processing is
149
+ required to get started).
150
+
151
+ ## Files
152
+
153
+ - `bioformbench_real.csv` — 49 rows, BioFormBench-Real (see schema above).
154
+ - `bioformbench_marketed.csv` — 165 rows, BioFormBench-Marketed (see schema
155
+ above; full computed-descriptor and label-parse columns included).
156
+
157
+ ## Known limitations
158
+
159
+ - BioFormBench-Real: n=13 proteins (9 LOPO-eligible) is small; treat any
160
+ per-protein result as illustrative, and prefer the aggregate statistics with
161
+ their reported confidence intervals.
162
+ - BioFormBench-Marketed: pH is extracted with a confidence flag
163
+ (`ph_is_range`/composition-sentence provenance in the full release columns);
164
+ the low-confidence stratum is retained rather than silently dropped, and the
165
+ companion paper shows the main finding (no sequence effect) holds in the
166
+ high-confidence stratum alone. Only 27–28 of 80 molecules have a fully
167
+ resolved buffer species *and* concentration; buffer-level analyses should
168
+ use that subset and are correspondingly lower-powered.
169
+ - Neither dataset should be used to generate or select actual formulations for
170
+ laboratory or clinical use without independent verification; see the
171
+ companion paper's Ethics statement.
172
+
173
+ ## Citation
174
+
175
+ ```bibtex
176
+ @dataset{kumar2026bioformbench,
177
+ title={BioFormBench: Two Open Benchmarks for Biologics Formulation Research},
178
+ author={Kumar, Bonthada Sravan},
179
+ year={2026},
180
+ publisher={Hugging Face Datasets},
181
+ url={https://huggingface.co/datasets/Sravankumarbonthada/BioFormBench}
182
+ }
183
+ ```
184
+
185
+ ## Sources
186
+
187
+ BioFormBench-Marketed is built from public FDA Structured Product Labeling
188
+ (DailyMed) and Thera-SAbDab (Raybould et al., Nucleic Acids Research 2020).
189
+ BioFormBench-Real is mined from 11 open-access PubMed Central articles, each
190
+ cited per-row via `source_id`.
191
+
192
+ ## Contact
193
+
194
+ Bonthada Sravan Kumar — sravansaijohn@gmail.com — issues via GitHub
195
+ (Maheshbonthada/bioform-lm)
196
+
197
+ ## License
198
+
199
+ CC-BY-4.0
200
+
201
+ ---
202
+
203
+ **Last updated:** September 8, 2026
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bioformbench_extracted_real_20260903.csv DELETED
@@ -1,9 +0,0 @@
1
- protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c,source_paper,source_doi_pmcid
2
- PGT121_acetate_baseline,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01}",50.0,0.50,50.0,-2.0,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
3
- PGT121_acetate_MoNi,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01, ""MoNi_copolymer"": 10.0}",50.0,0.70,30.0,1.5,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
4
- Adalimumab_commercial,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 50.0, ""polysorbate80"": 0.01}",25.0,0.95,5.0,0.0,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
5
- Adalimumab_concentrated_mannitol,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",4.0,0.98,2.0,0.5,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
6
- Adalimumab_concentrated_mannitol_RT,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",25.0,0.98,2.0,0.3,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
7
- X2Y1_bispecific_histidine,145.0,7.0,59.0,histidine,20.0,7.0,200.0,330.0,"{""sucrose"": 70.0, ""polysorbate80"": 0.01, ""arginine_hcl"": 200.0}",25.0,0.96,4.0,2.0,A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage,PMC11259756
8
- mAb2_citrate_baseline,150.0,7.2,65.0,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.85,15.0,-1.0,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
9
- mAb2_citrate_variant,150.0,7.2,66.5,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.92,8.0,1.5,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
 
 
 
 
 
 
 
 
 
 
bioformbench_extracted_real_batch2_20260903.csv DELETED
@@ -1,17 +0,0 @@
1
- protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c,source_paper,source_doi_pmcid
2
- Trastuzumab_histidine_reference,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{}",25.0,0.994,0.6,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
3
- Trastuzumab_arginine,150.0,7.2,68.6,histidine,10.0,6.5,150.0,310.0,"{""arginine_hcl"": 150.0}",25.0,1.00,0.0,-3.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
4
- Trastuzumab_glutamate,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{""sodium_glutamate"": 150.0}",25.0,0.994,0.6,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
5
- Omalizumab_histidine_reference,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{}",25.0,0.989,1.1,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
6
- Omalizumab_arginine,150.0,7.2,68.6,histidine,10.0,6.5,150.0,310.0,"{""arginine_hcl"": 150.0}",25.0,0.997,0.3,-3.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
7
- bispecific_A_histidine_no_salt,110.0,6.9,58.0,histidine,50.0,5.5,0.0,150.0,"{}",25.0,0.96,4.0,-2.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
8
- bispecific_A_histidine_150mM_NaCl,110.0,6.9,58.0,histidine,50.0,5.5,150.0,300.0,"{""nacl"": 150.0}",25.0,0.94,6.0,-1.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
9
- bispecific_A_histidine_500mM_NaCl,110.0,6.9,58.0,histidine,50.0,5.5,500.0,900.0,"{""nacl"": 500.0}",25.0,0.937,6.3,-1.2,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
10
- bispecific_A_histidine_Arg,110.0,6.9,58.0,histidine,50.0,5.5,150.0,300.0,"{""arginine_hcl"": 150.0}",25.0,0.941,5.9,-1.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
11
- bispecific_A_acetate,110.0,6.9,58.0,acetate,50.0,5.5,50.0,180.0,"{}",25.0,0.944,5.6,-1.5,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
12
- bispecific_A_citrate,110.0,6.9,58.0,citrate,50.0,5.5,50.0,180.0,"{}",25.0,0.961,3.9,-1.8,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
13
- bispecific_A_phosphate,110.0,6.9,58.0,phosphate,50.0,5.5,50.0,180.0,"{}",25.0,0.948,5.2,-1.4,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
14
- bispecific_A_histidine_40C,110.0,6.9,58.0,histidine,50.0,5.5,0.0,150.0,"{}",40.0,0.95,5.0,-1.5,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
15
- mAb_aggregation_prone_1,150.0,7.1,60.0,histidine,20.0,6.0,140.0,280.0,"{}",45.0,0.85,15.0,-0.5,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
16
- mAb_stable_reference,150.0,7.1,55.0,histidine,20.0,6.0,140.0,280.0,"{}",45.0,0.97,3.0,1.0,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
17
- mAb_high_viscosity_1,150.0,7.2,56.5,histidine,20.0,6.0,140.0,280.0,"{}",25.0,0.92,8.0,0.5,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bioformbench_extracted_real_batch3_20260903.csv DELETED
@@ -1,15 +0,0 @@
1
- protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c,source_paper,source_doi_pmcid
2
- A33_Fab_citrate_no_glycine,27.0,6.8,77.1,citrate,20.0,4.5,50.0,150.0,"{}",25.0,0.90,10.0,0.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
3
- A33_Fab_citrate_glycine10,27.0,6.8,79.1,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 10.0}",25.0,0.92,8.0,2.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
4
- A33_Fab_citrate_glycine20,27.0,6.8,79.6,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 20.0}",25.0,0.93,7.0,2.5,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
5
- A33_Fab_citrate_glycine30,27.0,6.8,82.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 30.0}",25.0,0.94,6.0,5.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
6
- A33_Fab_citrate_glycine40,27.0,6.8,82.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 40.0}",25.0,0.95,5.0,5.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
7
- A33_Fab_citrate_glycine50,27.0,6.8,83.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 50.0}",25.0,0.96,4.0,6.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
8
- A33_Fab_citrate_glycine60,27.0,6.8,83.6,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 60.0}",25.0,0.97,3.0,6.5,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
9
- A33_Fab_phosphate_glycine20,27.0,6.8,78.0,phosphate,10.0,7.0,50.0,150.0,"{""glycine"": 20.0}",25.0,0.94,6.0,1.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
10
- Fab_citrate_dilute,27.5,6.9,70.8,citrate,20.0,4.0,200.0,400.0,"{}",25.0,0.88,12.0,0.0,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
11
- Fab_citrate_medium,27.5,6.9,77.2,citrate,20.0,4.0,200.0,400.0,"{}",25.0,0.93,7.0,6.4,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
12
- Fab_citrate_concentrated,27.5,6.9,79.9,citrate,20.0,4.0,200.0,400.0,"{}",25.0,0.95,5.0,9.1,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
13
- Fab_citrate_high_conc_tween80,27.5,6.9,79.9,citrate,20.0,4.0,200.0,400.0,"{""tween80"": 0.01}",25.0,0.96,4.0,9.1,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
14
- Biosimilar_mAb_phosphate_cold,150.0,7.3,72.0,phosphate,20.0,7.4,188.0,380.0,"{}",5.0,0.98,2.0,0.0,Stability Evaluation of the Biosimilar Monoclonal Antibody Using Analytical Techniques,PMC9986942
15
- Biosimilar_mAb_phosphate_RT,150.0,7.3,72.0,phosphate,20.0,7.4,188.0,380.0,"{}",25.0,0.96,4.0,-2.0,Stability Evaluation of the Biosimilar Monoclonal Antibody Using Analytical Techniques,PMC9986942
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
bioformbench_extracted_real_batch4_20260903.csv DELETED
@@ -1,4 +0,0 @@
1
- protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c,source_paper,source_doi_pmcid
2
- rhIL1ra_MES_60mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,60.0,140.0,"{""NaCl"": 60.0}",25.0,0.85,15.0,0.0,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
3
- rhIL1ra_MES_150mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,150.0,300.0,"{""NaCl"": 150.0}",25.0,0.90,10.0,1.0,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
4
- rhIL1ra_MES_500mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,500.0,950.0,"{""NaCl"": 500.0}",25.0,0.94,6.0,2.5,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
 
 
 
 
 
bioformbench_marketed.csv ADDED
The diff for this file is too large to render. See raw diff
 
bioformbench_seed.csv → bioformbench_real.csv RENAMED
@@ -1,55 +1,37 @@
1
- protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c
2
- protein_1,50.0,7.2,65.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0}",25.0,0.85,2.5,8.0
3
- protein_1,50.0,7.2,65.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.92,1.2,12.0
4
- protein_1,50.0,7.2,65.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.65,5.8,4.0
5
- protein_1,50.0,7.2,65.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0, ""tween80"": 0.01}",25.0,0.88,1.8,10.0
6
- protein_2,75.0,6.8,58.0,histidine,50.0,6.5,150.0,300.0,"{""sucrose"": 100.0}",25.0,0.80,3.5,6.0
7
- protein_2,75.0,6.8,58.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.89,1.5,11.0
8
- protein_2,75.0,6.8,58.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.72,4.2,5.0
9
- protein_2,75.0,6.8,58.0,histidine,50.0,6.5,150.0,300.0,"{""sucrose"": 100.0, ""polysorbate20"": 0.01}",25.0,0.83,2.8,7.5
10
- protein_3,45.0,8.1,62.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0}",25.0,0.82,3.2,7.5
11
- protein_3,45.0,8.1,62.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 250.0}",25.0,0.90,1.0,13.0
12
- protein_3,45.0,8.1,62.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.68,5.5,3.5
13
- protein_3,45.0,8.1,62.0,citrate,60.0,5.0,120.0,280.0,"{""glycerol"": 10.0}",25.0,0.75,4.0,6.0
14
- protein_4,120.0,5.9,70.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0}",25.0,0.84,2.8,8.5
15
- protein_4,120.0,5.9,70.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.91,0.9,12.5
16
- protein_4,120.0,5.9,70.0,acetate,75.0,5.5,100.0,250.0,"{""mannitol"": 150.0}",25.0,0.70,4.5,4.5
17
- protein_5,60.0,7.5,64.0,histidine,50.0,6.5,150.0,300.0,"{""sucrose"": 100.0}",25.0,0.81,3.5,7.0
18
- protein_5,60.0,7.5,64.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.88,1.3,11.0
19
- protein_5,60.0,7.5,64.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.69,4.8,4.8
20
- PGT121_acetate_baseline,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01}",50.0,0.50,50.0,-2.0,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
21
- PGT121_acetate_MoNi,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01, ""MoNi_copolymer"": 10.0}",50.0,0.70,30.0,1.5,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
22
  Adalimumab_commercial,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 50.0, ""polysorbate80"": 0.01}",25.0,0.95,5.0,0.0,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
23
  Adalimumab_concentrated_mannitol,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",4.0,0.98,2.0,0.5,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
24
  Adalimumab_concentrated_mannitol_RT,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",25.0,0.98,2.0,0.3,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
25
  X2Y1_bispecific_histidine,145.0,7.0,59.0,histidine,20.0,7.0,200.0,330.0,"{""sucrose"": 70.0, ""polysorbate80"": 0.01, ""arginine_hcl"": 200.0}",25.0,0.96,4.0,2.0,A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage,PMC11259756
26
- mAb2_citrate_baseline,150.0,7.2,65.0,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.85,15.0,-1.0,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
27
- mAb2_citrate_variant,150.0,7.2,66.5,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.92,8.0,1.5,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
28
- PGT121_acetate_baseline,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01}",50.0,0.50,50.0,-2.0,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
29
- PGT121_acetate_MoNi,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01, ""MoNi_copolymer"": 10.0}",50.0,0.70,30.0,1.5,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
30
  Adalimumab_commercial,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 50.0, ""polysorbate80"": 0.01}",25.0,0.95,5.0,0.0,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
31
  Adalimumab_concentrated_mannitol,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",4.0,0.98,2.0,0.5,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
32
  Adalimumab_concentrated_mannitol_RT,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",25.0,0.98,2.0,0.3,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
33
  X2Y1_bispecific_histidine,145.0,7.0,59.0,histidine,20.0,7.0,200.0,330.0,"{""sucrose"": 70.0, ""polysorbate80"": 0.01, ""arginine_hcl"": 200.0}",25.0,0.96,4.0,2.0,A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage,PMC11259756
34
- mAb2_citrate_baseline,150.0,7.2,65.0,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.85,15.0,-1.0,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
35
- mAb2_citrate_variant,150.0,7.2,66.5,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.92,8.0,1.5,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
36
- Trastuzumab_histidine_reference,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{}",25.0,0.994,0.6,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
37
- Trastuzumab_arginine,150.0,7.2,68.6,histidine,10.0,6.5,150.0,310.0,"{""arginine_hcl"": 150.0}",25.0,1.00,0.0,-3.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
38
  Trastuzumab_glutamate,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{""sodium_glutamate"": 150.0}",25.0,0.994,0.6,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
39
- Omalizumab_histidine_reference,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{}",25.0,0.989,1.1,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
40
  Omalizumab_arginine,150.0,7.2,68.6,histidine,10.0,6.5,150.0,310.0,"{""arginine_hcl"": 150.0}",25.0,0.997,0.3,-3.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
41
- bispecific_A_histidine_no_salt,110.0,6.9,58.0,histidine,50.0,5.5,0.0,150.0,"{}",25.0,0.96,4.0,-2.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
42
  bispecific_A_histidine_150mM_NaCl,110.0,6.9,58.0,histidine,50.0,5.5,150.0,300.0,"{""nacl"": 150.0}",25.0,0.94,6.0,-1.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
43
  bispecific_A_histidine_500mM_NaCl,110.0,6.9,58.0,histidine,50.0,5.5,500.0,900.0,"{""nacl"": 500.0}",25.0,0.937,6.3,-1.2,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
44
  bispecific_A_histidine_Arg,110.0,6.9,58.0,histidine,50.0,5.5,150.0,300.0,"{""arginine_hcl"": 150.0}",25.0,0.941,5.9,-1.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
45
- bispecific_A_acetate,110.0,6.9,58.0,acetate,50.0,5.5,50.0,180.0,"{}",25.0,0.944,5.6,-1.5,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
46
- bispecific_A_citrate,110.0,6.9,58.0,citrate,50.0,5.5,50.0,180.0,"{}",25.0,0.961,3.9,-1.8,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
47
- bispecific_A_phosphate,110.0,6.9,58.0,phosphate,50.0,5.5,50.0,180.0,"{}",25.0,0.948,5.2,-1.4,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
48
- bispecific_A_histidine_40C,110.0,6.9,58.0,histidine,50.0,5.5,0.0,150.0,"{}",40.0,0.95,5.0,-1.5,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
49
- mAb_aggregation_prone_1,150.0,7.1,60.0,histidine,20.0,6.0,140.0,280.0,"{}",45.0,0.85,15.0,-0.5,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
50
- mAb_stable_reference,150.0,7.1,55.0,histidine,20.0,6.0,140.0,280.0,"{}",45.0,0.97,3.0,1.0,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
51
- mAb_high_viscosity_1,150.0,7.2,56.5,histidine,20.0,6.0,140.0,280.0,"{}",25.0,0.92,8.0,0.5,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
52
- A33_Fab_citrate_no_glycine,27.0,6.8,77.1,citrate,20.0,4.5,50.0,150.0,"{}",25.0,0.90,10.0,0.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
53
  A33_Fab_citrate_glycine10,27.0,6.8,79.1,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 10.0}",25.0,0.92,8.0,2.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
54
  A33_Fab_citrate_glycine20,27.0,6.8,79.6,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 20.0}",25.0,0.93,7.0,2.5,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
55
  A33_Fab_citrate_glycine30,27.0,6.8,82.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 30.0}",25.0,0.94,6.0,5.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
@@ -57,12 +39,12 @@ A33_Fab_citrate_glycine40,27.0,6.8,82.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"
57
  A33_Fab_citrate_glycine50,27.0,6.8,83.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 50.0}",25.0,0.96,4.0,6.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
58
  A33_Fab_citrate_glycine60,27.0,6.8,83.6,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 60.0}",25.0,0.97,3.0,6.5,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
59
  A33_Fab_phosphate_glycine20,27.0,6.8,78.0,phosphate,10.0,7.0,50.0,150.0,"{""glycine"": 20.0}",25.0,0.94,6.0,1.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
60
- Fab_citrate_dilute,27.5,6.9,70.8,citrate,20.0,4.0,200.0,400.0,"{}",25.0,0.88,12.0,0.0,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
61
- Fab_citrate_medium,27.5,6.9,77.2,citrate,20.0,4.0,200.0,400.0,"{}",25.0,0.93,7.0,6.4,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
62
- Fab_citrate_concentrated,27.5,6.9,79.9,citrate,20.0,4.0,200.0,400.0,"{}",25.0,0.95,5.0,9.1,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
63
  Fab_citrate_high_conc_tween80,27.5,6.9,79.9,citrate,20.0,4.0,200.0,400.0,"{""tween80"": 0.01}",25.0,0.96,4.0,9.1,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
64
- Biosimilar_mAb_phosphate_cold,150.0,7.3,72.0,phosphate,20.0,7.4,188.0,380.0,"{}",5.0,0.98,2.0,0.0,Stability Evaluation of the Biosimilar Monoclonal Antibody Using Analytical Techniques,PMC9986942
65
- Biosimilar_mAb_phosphate_RT,150.0,7.3,72.0,phosphate,20.0,7.4,188.0,380.0,"{}",25.0,0.96,4.0,-2.0,Stability Evaluation of the Biosimilar Monoclonal Antibody Using Analytical Techniques,PMC9986942
66
  rhIL1ra_MES_60mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,60.0,140.0,"{""NaCl"": 60.0}",25.0,0.85,15.0,0.0,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
67
- rhIL1ra_MES_150mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,150.0,300.0,"{""NaCl"": 150.0}",25.0,0.90,10.0,1.0,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
68
  rhIL1ra_MES_500mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,500.0,950.0,"{""NaCl"": 500.0}",25.0,0.94,6.0,2.5,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
 
1
+ protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c,source_title,source_id
2
+ PGT121_acetate_baseline,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01}",50.0,0.5,50.0,-2.0,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
3
+ PGT121_acetate_MoNi,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01, ""MoNi_copolymer"": 10.0}",50.0,0.7,30.0,1.5,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4
  Adalimumab_commercial,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 50.0, ""polysorbate80"": 0.01}",25.0,0.95,5.0,0.0,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
5
  Adalimumab_concentrated_mannitol,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",4.0,0.98,2.0,0.5,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
6
  Adalimumab_concentrated_mannitol_RT,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",25.0,0.98,2.0,0.3,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
7
  X2Y1_bispecific_histidine,145.0,7.0,59.0,histidine,20.0,7.0,200.0,330.0,"{""sucrose"": 70.0, ""polysorbate80"": 0.01, ""arginine_hcl"": 200.0}",25.0,0.96,4.0,2.0,A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage,PMC11259756
8
+ mAb2_citrate_baseline,150.0,7.2,65.0,citrate,25.0,6.0,165.0,300.0,{},25.0,0.85,15.0,-1.0,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
9
+ mAb2_citrate_variant,150.0,7.2,66.5,citrate,25.0,6.0,165.0,300.0,{},25.0,0.92,8.0,1.5,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
10
+ PGT121_acetate_baseline,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01}",50.0,0.5,50.0,-2.0,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
11
+ PGT121_acetate_MoNi,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01, ""MoNi_copolymer"": 10.0}",50.0,0.7,30.0,1.5,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
12
  Adalimumab_commercial,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 50.0, ""polysorbate80"": 0.01}",25.0,0.95,5.0,0.0,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
13
  Adalimumab_concentrated_mannitol,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",4.0,0.98,2.0,0.5,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
14
  Adalimumab_concentrated_mannitol_RT,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",25.0,0.98,2.0,0.3,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
15
  X2Y1_bispecific_histidine,145.0,7.0,59.0,histidine,20.0,7.0,200.0,330.0,"{""sucrose"": 70.0, ""polysorbate80"": 0.01, ""arginine_hcl"": 200.0}",25.0,0.96,4.0,2.0,A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage,PMC11259756
16
+ mAb2_citrate_baseline,150.0,7.2,65.0,citrate,25.0,6.0,165.0,300.0,{},25.0,0.85,15.0,-1.0,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
17
+ mAb2_citrate_variant,150.0,7.2,66.5,citrate,25.0,6.0,165.0,300.0,{},25.0,0.92,8.0,1.5,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
18
+ Trastuzumab_histidine_reference,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,{},25.0,0.994,0.6,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
19
+ Trastuzumab_arginine,150.0,7.2,68.6,histidine,10.0,6.5,150.0,310.0,"{""arginine_hcl"": 150.0}",25.0,1.0,0.0,-3.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
20
  Trastuzumab_glutamate,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,"{""sodium_glutamate"": 150.0}",25.0,0.994,0.6,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
21
+ Omalizumab_histidine_reference,150.0,7.2,71.6,histidine,10.0,6.5,150.0,310.0,{},25.0,0.989,1.1,0.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
22
  Omalizumab_arginine,150.0,7.2,68.6,histidine,10.0,6.5,150.0,310.0,"{""arginine_hcl"": 150.0}",25.0,0.997,0.3,-3.0,Residue-resolved insights into the stabilization of therapeutic proteins by excipients,PMC11572152
23
+ bispecific_A_histidine_no_salt,110.0,6.9,58.0,histidine,50.0,5.5,0.0,150.0,{},25.0,0.96,4.0,-2.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
24
  bispecific_A_histidine_150mM_NaCl,110.0,6.9,58.0,histidine,50.0,5.5,150.0,300.0,"{""nacl"": 150.0}",25.0,0.94,6.0,-1.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
25
  bispecific_A_histidine_500mM_NaCl,110.0,6.9,58.0,histidine,50.0,5.5,500.0,900.0,"{""nacl"": 500.0}",25.0,0.937,6.3,-1.2,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
26
  bispecific_A_histidine_Arg,110.0,6.9,58.0,histidine,50.0,5.5,150.0,300.0,"{""arginine_hcl"": 150.0}",25.0,0.941,5.9,-1.0,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
27
+ bispecific_A_acetate,110.0,6.9,58.0,acetate,50.0,5.5,50.0,180.0,{},25.0,0.944,5.6,-1.5,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
28
+ bispecific_A_citrate,110.0,6.9,58.0,citrate,50.0,5.5,50.0,180.0,{},25.0,0.961,3.9,-1.8,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
29
+ bispecific_A_phosphate,110.0,6.9,58.0,phosphate,50.0,5.5,50.0,180.0,{},25.0,0.948,5.2,-1.4,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
30
+ bispecific_A_histidine_40C,110.0,6.9,58.0,histidine,50.0,5.5,0.0,150.0,{},40.0,0.95,5.0,-1.5,Investigation on environmental factors contributing to bispecific antibody stability,PMC11343937
31
+ mAb_aggregation_prone_1,150.0,7.1,60.0,histidine,20.0,6.0,140.0,280.0,{},45.0,0.85,15.0,-0.5,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
32
+ mAb_stable_reference,150.0,7.1,55.0,histidine,20.0,6.0,140.0,280.0,{},45.0,0.97,3.0,1.0,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
33
+ mAb_high_viscosity_1,150.0,7.2,56.5,histidine,20.0,6.0,140.0,280.0,{},25.0,0.92,8.0,0.5,Machine learning prediction of antibody aggregation and viscosity,PMC8794240
34
+ A33_Fab_citrate_no_glycine,27.0,6.8,77.1,citrate,20.0,4.5,50.0,150.0,{},25.0,0.9,10.0,0.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
35
  A33_Fab_citrate_glycine10,27.0,6.8,79.1,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 10.0}",25.0,0.92,8.0,2.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
36
  A33_Fab_citrate_glycine20,27.0,6.8,79.6,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 20.0}",25.0,0.93,7.0,2.5,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
37
  A33_Fab_citrate_glycine30,27.0,6.8,82.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 30.0}",25.0,0.94,6.0,5.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
 
39
  A33_Fab_citrate_glycine50,27.0,6.8,83.2,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 50.0}",25.0,0.96,4.0,6.1,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
40
  A33_Fab_citrate_glycine60,27.0,6.8,83.6,citrate,20.0,4.5,50.0,150.0,"{""glycine"": 60.0}",25.0,0.97,3.0,6.5,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
41
  A33_Fab_phosphate_glycine20,27.0,6.8,78.0,phosphate,10.0,7.0,50.0,150.0,"{""glycine"": 20.0}",25.0,0.94,6.0,1.0,Molecular Dynamics Simulations Reveal Protein-Surface Interactions Modulate Stability in Fab Formulations,PMC11539065
42
+ Fab_citrate_dilute,27.5,6.9,70.8,citrate,20.0,4.0,200.0,400.0,{},25.0,0.88,12.0,0.0,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
43
+ Fab_citrate_medium,27.5,6.9,77.2,citrate,20.0,4.0,200.0,400.0,{},25.0,0.93,7.0,6.4,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
44
+ Fab_citrate_concentrated,27.5,6.9,79.9,citrate,20.0,4.0,200.0,400.0,{},25.0,0.95,5.0,9.1,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
45
  Fab_citrate_high_conc_tween80,27.5,6.9,79.9,citrate,20.0,4.0,200.0,400.0,"{""tween80"": 0.01}",25.0,0.96,4.0,9.1,Enhanced Thermal Stability and Reduced Aggregation in Fab Fragment at Elevated Concentrations,PMC10155210
46
+ Biosimilar_mAb_phosphate_cold,150.0,7.3,72.0,phosphate,20.0,7.4,188.0,380.0,{},5.0,0.98,2.0,0.0,Stability Evaluation of the Biosimilar Monoclonal Antibody Using Analytical Techniques,PMC9986942
47
+ Biosimilar_mAb_phosphate_RT,150.0,7.3,72.0,phosphate,20.0,7.4,188.0,380.0,{},25.0,0.96,4.0,-2.0,Stability Evaluation of the Biosimilar Monoclonal Antibody Using Analytical Techniques,PMC9986942
48
  rhIL1ra_MES_60mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,60.0,140.0,"{""NaCl"": 60.0}",25.0,0.85,15.0,0.0,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
49
+ rhIL1ra_MES_150mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,150.0,300.0,"{""NaCl"": 150.0}",25.0,0.9,10.0,1.0,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
50
  rhIL1ra_MES_500mM_NaCl,17.3,5.6,55.0,MES,10.0,6.5,500.0,950.0,"{""NaCl"": 500.0}",25.0,0.94,6.0,2.5,Protein-protein interactions controlling interfacial aggregation of rhIL-1ra,PMC6032359
bioformbench_seed_backup_20260903.csv DELETED
@@ -1,27 +0,0 @@
1
- protein_id,protein_mw_kda,protein_pi,protein_tm_baseline_c,buffer_species,buffer_conc_mm,ph,ionic_strength_mm,osmolarity_mosm_kg,stabilizers_json,temperature_c,stability_score,measured_aggregation_percent,measured_tm_shift_c
2
- protein_1,50.0,7.2,65.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0}",25.0,0.85,2.5,8.0
3
- protein_1,50.0,7.2,65.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.92,1.2,12.0
4
- protein_1,50.0,7.2,65.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.65,5.8,4.0
5
- protein_1,50.0,7.2,65.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0, ""tween80"": 0.01}",25.0,0.88,1.8,10.0
6
- protein_2,75.0,6.8,58.0,histidine,50.0,6.5,150.0,300.0,"{""sucrose"": 100.0}",25.0,0.80,3.5,6.0
7
- protein_2,75.0,6.8,58.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.89,1.5,11.0
8
- protein_2,75.0,6.8,58.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.72,4.2,5.0
9
- protein_2,75.0,6.8,58.0,histidine,50.0,6.5,150.0,300.0,"{""sucrose"": 100.0, ""polysorbate20"": 0.01}",25.0,0.83,2.8,7.5
10
- protein_3,45.0,8.1,62.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0}",25.0,0.82,3.2,7.5
11
- protein_3,45.0,8.1,62.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 250.0}",25.0,0.90,1.0,13.0
12
- protein_3,45.0,8.1,62.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.68,5.5,3.5
13
- protein_3,45.0,8.1,62.0,citrate,60.0,5.0,120.0,280.0,"{""glycerol"": 10.0}",25.0,0.75,4.0,6.0
14
- protein_4,120.0,5.9,70.0,histidine,50.0,6.5,150.0,300.0,"{""trehalose"": 100.0}",25.0,0.84,2.8,8.5
15
- protein_4,120.0,5.9,70.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.91,0.9,12.5
16
- protein_4,120.0,5.9,70.0,acetate,75.0,5.5,100.0,250.0,"{""mannitol"": 150.0}",25.0,0.70,4.5,4.5
17
- protein_5,60.0,7.5,64.0,histidine,50.0,6.5,150.0,300.0,"{""sucrose"": 100.0}",25.0,0.81,3.5,7.0
18
- protein_5,60.0,7.5,64.0,phosphate,100.0,7.0,200.0,350.0,"{""trehalose"": 200.0}",25.0,0.88,1.3,11.0
19
- protein_5,60.0,7.5,64.0,acetate,75.0,5.5,100.0,250.0,"{""sorbitol"": 150.0}",25.0,0.69,4.8,4.8
20
- PGT121_acetate_baseline,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01}",50.0,0.50,50.0,-2.0,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
21
- PGT121_acetate_MoNi,146.0,7.0,60.0,acetate,20.0,5.2,50.0,150.0,"{""sucrose"": 90.0, ""polysorbate80"": 0.01, ""MoNi_copolymer"": 10.0}",50.0,0.70,30.0,1.5,Stable High-Concentration Monoclonal Antibody Formulations Enabled by an Amphiphilic Copolymer Excipient,PMC9854243
22
- Adalimumab_commercial,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 50.0, ""polysorbate80"": 0.01}",25.0,0.95,5.0,0.0,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
23
- Adalimumab_concentrated_mannitol,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",4.0,0.98,2.0,0.5,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
24
- Adalimumab_concentrated_mannitol_RT,148.0,7.2,72.0,citrate,20.0,5.3,140.0,280.0,"{""mannitol"": 12.0, ""polysorbate80"": 0.01}",25.0,0.98,2.0,0.3,Stability of Monoclonal Antibodies as Solid Formulation for Auto-Injectors: A Pilot Study,PMC10459033
25
- X2Y1_bispecific_histidine,145.0,7.0,59.0,histidine,20.0,7.0,200.0,330.0,"{""sucrose"": 70.0, ""polysorbate80"": 0.01, ""arginine_hcl"": 200.0}",25.0,0.96,4.0,2.0,A case study of a bispecific antibody manufacturability assessment and optimization during discovery stage,PMC11259756
26
- mAb2_citrate_baseline,150.0,7.2,65.0,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.85,15.0,-1.0,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
27
- mAb2_citrate_variant,150.0,7.2,66.5,citrate,25.0,6.0,165.0,300.0,"{}",25.0,0.92,8.0,1.5,Rational optimization of a monoclonal antibody improves the aggregation propensity and enhances the CMC properties,PMC7531517
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
synthetic_generator.py DELETED
@@ -1,363 +0,0 @@
1
- """
2
- Synthetic Formulation Data Generator
3
-
4
- Generates high-quality synthetic (protein, formulation) → (stability_outcome) triples
5
- using the validated mechanistic simulator. This synthetic corpus is used for
6
- pretraining the transformer model.
7
-
8
- Production features:
9
- - Stratified sampling (covers diverse protein/formulation space)
10
- - Quality filtering (removes unphysical samples)
11
- - Reproducible with seed control
12
- - Efficient batch generation
13
- - Parallel generation capability
14
- """
15
-
16
- import numpy as np
17
- import pandas as pd
18
- from typing import List, Dict, Tuple, Optional
19
- from pathlib import Path
20
- import logging
21
- from dataclasses import dataclass
22
- import json
23
- from tqdm import tqdm
24
- import sys
25
-
26
- sys.path.insert(0, str(Path(__file__).parent.parent))
27
-
28
- from simulator.mechanistic_sim import (
29
- BiologicsFormulationSimulator,
30
- ProteinDescriptor,
31
- FormulationComposition,
32
- SimulationOutput,
33
- )
34
- from config import CFG, SyntheticDataConfig
35
-
36
- logger = logging.getLogger(__name__)
37
-
38
-
39
- @dataclass
40
- class SyntheticSample:
41
- """Single synthetic training sample."""
42
- protein_mw_kda: float
43
- protein_pi: float
44
- protein_hydrophobicity: float
45
- protein_tm_baseline_c: float
46
- buffer_species: str
47
- buffer_conc_mm: float
48
- ph: float
49
- ionic_strength_mm: float
50
- osmolarity_mosm_kg: float
51
- stabilizers_json: str # JSON string of stabilizer dict
52
- temperature_c: float
53
- storage_duration_days: float
54
- # Outcomes
55
- tm_shift_c: float
56
- predicted_tm_c: float
57
- aggregation_risk_combined: float
58
- predicted_aggregation_percent: float
59
- stability_score: float
60
-
61
- def to_dict(self) -> Dict:
62
- """Convert to dict for DataFrame."""
63
- return {
64
- 'protein_mw_kda': self.protein_mw_kda,
65
- 'protein_pi': self.protein_pi,
66
- 'protein_hydrophobicity': self.protein_hydrophobicity,
67
- 'protein_tm_baseline_c': self.protein_tm_baseline_c,
68
- 'buffer_species': self.buffer_species,
69
- 'buffer_conc_mm': self.buffer_conc_mm,
70
- 'ph': self.ph,
71
- 'ionic_strength_mm': self.ionic_strength_mm,
72
- 'osmolarity_mosm_kg': self.osmolarity_mosm_kg,
73
- 'stabilizers_json': self.stabilizers_json,
74
- 'temperature_c': self.temperature_c,
75
- 'storage_duration_days': self.storage_duration_days,
76
- 'tm_shift_c': self.tm_shift_c,
77
- 'predicted_tm_c': self.predicted_tm_c,
78
- 'aggregation_risk_combined': self.aggregation_risk_combined,
79
- 'predicted_aggregation_percent': self.predicted_aggregation_percent,
80
- 'stability_score': self.stability_score,
81
- }
82
-
83
-
84
- class SyntheticDataGenerator:
85
- """Generates high-quality synthetic training data."""
86
-
87
- def __init__(
88
- self,
89
- num_samples: int = 500_000,
90
- seed: int = 42,
91
- config: Optional[SyntheticDataConfig] = None,
92
- ):
93
- """
94
- Initialize synthetic data generator.
95
-
96
- Args:
97
- num_samples: Number of synthetic samples to generate
98
- seed: Random seed for reproducibility
99
- config: SyntheticDataConfig from config.py
100
- """
101
- self.num_samples = num_samples
102
- self.seed = seed
103
- self.rng = np.random.RandomState(seed)
104
- self.config = config or CFG.synthetic_data
105
-
106
- # Initialize simulator
107
- self.simulator = BiologicsFormulationSimulator(
108
- config_dict={
109
- 'hamaker_constant': CFG.simulator.hamaker_constant,
110
- 'bjerrum_length': CFG.simulator.bjerrum_length,
111
- 'ph_penalty_per_unit_sq': CFG.simulator.ph_penalty_per_unit_sq,
112
- 'osmol_positive_slope': CFG.simulator.osmol_positive_slope,
113
- 'osmol_negative_slope': CFG.simulator.osmol_negative_slope,
114
- 'aggregation_rate_constant': CFG.simulator.aggregation_rate_constant,
115
- 'stabilizer_effects': CFG.simulator.stabilizer_effects,
116
- }
117
- )
118
-
119
- logger.info(f"SyntheticDataGenerator initialized: {num_samples} samples, seed={seed}")
120
-
121
- def generate(self) -> pd.DataFrame:
122
- """
123
- Generate synthetic dataset.
124
-
125
- Returns:
126
- DataFrame with synthetic (protein, formulation) → outcome triples
127
-
128
- Process:
129
- 1. Sample random proteins from descriptor space
130
- 2. Sample random formulations from composition space
131
- 3. Run simulator for each pair
132
- 4. Filter out unphysical/invalid samples
133
- 5. Return as DataFrame
134
-
135
- This is CPU-efficient and can generate millions of samples.
136
- """
137
- samples = []
138
- skipped = 0
139
-
140
- logger.info(f"Generating {self.num_samples} synthetic samples...")
141
-
142
- for i in tqdm(range(self.num_samples), desc="Generating synthetic data"):
143
- try:
144
- # Sample random protein
145
- protein = self._sample_protein()
146
-
147
- # Sample random formulation
148
- formulation = self._sample_formulation()
149
-
150
- # Run simulator
151
- output = self.simulator.predict(protein, formulation)
152
-
153
- # Create sample
154
- stabilizers_dict = formulation.stabilizers or {}
155
- sample = SyntheticSample(
156
- protein_mw_kda=protein.mw_kda,
157
- protein_pi=protein.pi,
158
- protein_hydrophobicity=protein.hydrophobicity,
159
- protein_tm_baseline_c=protein.tm_baseline_c,
160
- buffer_species=formulation.buffer_species,
161
- buffer_conc_mm=formulation.buffer_conc_mm,
162
- ph=formulation.ph,
163
- ionic_strength_mm=formulation.ionic_strength_mm,
164
- osmolarity_mosm_kg=formulation.osmolarity_mosm_kg,
165
- stabilizers_json=json.dumps(stabilizers_dict),
166
- temperature_c=formulation.temperature_c,
167
- storage_duration_days=formulation.storage_duration_days,
168
- tm_shift_c=output.tm_shift_c,
169
- predicted_tm_c=output.predicted_tm_c,
170
- aggregation_risk_combined=output.aggregation_risk_combined,
171
- predicted_aggregation_percent=output.predicted_aggregation_percent,
172
- stability_score=output.stability_score,
173
- )
174
-
175
- # Quality check: ensure sample is physically plausible
176
- if self._is_valid_sample(sample):
177
- samples.append(sample.to_dict())
178
- else:
179
- skipped += 1
180
-
181
- except Exception as e:
182
- skipped += 1
183
- if i % 10000 == 0:
184
- logger.debug(f"Skipped sample {i}: {e}")
185
- continue
186
-
187
- logger.info(f"Generated {len(samples)} valid samples, skipped {skipped}")
188
-
189
- # Convert to DataFrame
190
- df = pd.DataFrame(samples)
191
- return df
192
-
193
- def _sample_protein(self) -> ProteinDescriptor:
194
- """Sample random protein descriptor."""
195
- mw = self.rng.uniform(*self.config.protein_mw_range)
196
- pi = self.rng.uniform(*self.config.protein_pi_range)
197
- hydro = self.rng.uniform(*self.config.protein_hydrophobicity_range)
198
- tm = self.rng.uniform(30, 85) # typical Tm range
199
-
200
- return ProteinDescriptor(
201
- mw_kda=mw,
202
- pi=pi,
203
- hydrophobicity=hydro,
204
- tm_baseline_c=tm,
205
- )
206
-
207
- def _sample_formulation(self) -> FormulationComposition:
208
- """Sample random formulation composition."""
209
- # Buffer
210
- buffer_species = self.rng.choice(self.config.buffer_types)
211
- buffer_conc = self.rng.uniform(*self.config.buffer_concentration_range)
212
-
213
- # pH
214
- ph = self.rng.uniform(*self.config.ph_range)
215
-
216
- # Ionic strength
217
- ionic_strength = self.rng.uniform(*self.config.ionic_strength_range)
218
-
219
- # Osmolarity (derived from buffer + ions)
220
- osmolarity = self.rng.uniform(*self.config.osmolarity_range)
221
-
222
- # Stabilizers (multi-select)
223
- stabilizers = {}
224
- num_stab = self.rng.randint(0, self.config.num_stabilizers_per_sample + 1)
225
-
226
- for _ in range(num_stab):
227
- stab_name = self.rng.choice(list(self.config.stabilizer_pool.keys()))
228
- min_conc, max_conc = self.config.stabilizer_pool[stab_name]
229
-
230
- # Avoid adding zero-concentration stabilizers
231
- if max_conc > 0:
232
- conc = self.rng.uniform(min_conc + 0.01, max_conc)
233
- stabilizers[stab_name] = conc
234
-
235
- # Storage conditions
236
- temperature = self.rng.uniform(*self.config.temperature_range)
237
- duration = self.rng.uniform(*self.config.storage_duration_range)
238
-
239
- return FormulationComposition(
240
- buffer_species=buffer_species,
241
- buffer_conc_mm=buffer_conc,
242
- ph=ph,
243
- ionic_strength_mm=ionic_strength,
244
- osmolarity_mosm_kg=osmolarity,
245
- stabilizers=stabilizers,
246
- temperature_c=temperature,
247
- storage_duration_days=duration,
248
- )
249
-
250
- def _is_valid_sample(self, sample: SyntheticSample) -> bool:
251
- """
252
- Quality filter: reject unphysical or invalid samples.
253
-
254
- Checks:
255
- - Aggregation % in reasonable range
256
- - Stability score is valid
257
- - Outcomes are finite (not NaN/inf)
258
- """
259
- # Check for NaN/inf
260
- if not np.isfinite(sample.aggregation_risk_combined):
261
- return False
262
- if not np.isfinite(sample.stability_score):
263
- return False
264
-
265
- # Aggregation should be 0-100%
266
- if not (0 <= sample.predicted_aggregation_percent <= 100):
267
- return False
268
-
269
- # Stability score should be 0-1
270
- if not (0 <= sample.stability_score <= 1):
271
- return False
272
-
273
- # At least some buffer
274
- if sample.buffer_conc_mm < 1:
275
- return False
276
-
277
- return True
278
-
279
- def save(self, path: Path) -> None:
280
- """Generate and save synthetic data to CSV/Parquet."""
281
- df = self.generate()
282
-
283
- # Save as both CSV (human-readable) and Parquet (efficient)
284
- csv_path = path.with_suffix('.csv')
285
- parquet_path = path.with_suffix('.parquet')
286
-
287
- logger.info(f"Saving {len(df)} samples to {csv_path}")
288
- df.to_csv(csv_path, index=False)
289
-
290
- logger.info(f"Saving to Parquet: {parquet_path}")
291
- df.to_parquet(parquet_path, index=False)
292
-
293
- # Summary statistics
294
- logger.info("\n" + "=" * 70)
295
- logger.info("SYNTHETIC DATA SUMMARY")
296
- logger.info("=" * 70)
297
- logger.info(f"Total samples: {len(df)}")
298
- logger.info(f"\nProtein descriptors:")
299
- logger.info(df[['protein_mw_kda', 'protein_pi', 'protein_tm_baseline_c']].describe())
300
- logger.info(f"\nFormulation parameters:")
301
- logger.info(df[['ph', 'ionic_strength_mm', 'temperature_c']].describe())
302
- logger.info(f"\nOutcomes:")
303
- logger.info(df[['stability_score', 'predicted_aggregation_percent']].describe())
304
- logger.info("=" * 70)
305
-
306
- return df
307
-
308
-
309
- def generate_synthetic_dataset(
310
- output_dir: Path = None,
311
- num_samples: int = 500_000,
312
- seed: int = 42,
313
- ) -> pd.DataFrame:
314
- """
315
- Convenience function: generate and save synthetic dataset.
316
-
317
- Args:
318
- output_dir: Directory to save dataset (default: data/)
319
- num_samples: Number of samples (default: 500K)
320
- seed: Random seed (default: 42)
321
-
322
- Returns:
323
- Generated DataFrame
324
- """
325
- if output_dir is None:
326
- output_dir = Path(__file__).parent
327
-
328
- output_dir.mkdir(parents=True, exist_ok=True)
329
-
330
- # Generate
331
- generator = SyntheticDataGenerator(
332
- num_samples=num_samples,
333
- seed=seed,
334
- )
335
-
336
- # Save
337
- output_path = output_dir / "synthetic_training_data"
338
- df = generator.generate()
339
- generator.save(output_path)
340
-
341
- return df
342
-
343
-
344
- if __name__ == "__main__":
345
- """Generate synthetic dataset."""
346
- import sys
347
- from pathlib import Path
348
-
349
- logging.basicConfig(
350
- level=logging.INFO,
351
- format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
352
- )
353
-
354
- # Command line: python synthetic_generator.py <num_samples>
355
- num_samples = int(sys.argv[1]) if len(sys.argv) > 1 else 10_000
356
-
357
- df = generate_synthetic_dataset(
358
- num_samples=num_samples,
359
- seed=42,
360
- )
361
-
362
- print(f"\n✓ Generated {len(df)} synthetic samples")
363
- print(f"Dataset saved to: data/synthetic_training_data.*")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
synthetic_training_data.csv DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:1ac5efd796179d568251b835650ecf54f37681f3a771a15fc6b0081baa109173
3
- size 33722233
 
 
 
 
synthetic_training_data.parquet DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:a92c824956de1a8d877a2f28a4d89516ee0453d5c0d888921725da4fcc3fc4be
3
- size 17981929