Generated: 2026-02-08 (revised) Pipeline: Automated Green Chemistry Redesign Agent v1.0 Revision: Includes reversion risk analysis and corrected rankings
This report presents the results of an automated computational pipeline for redesigning carbamazepine (CBZ) with improved environmental fate while maintaining pharmacological activity. Starting from the parent anticonvulsant, the pipeline generated 27 structural modifications across three strategies: biodegradable linker insertion, polar group addition, and prodrug design.
Key results:
- 20 of 27 compounds (74%) are novel — not previously reported in PubChem or literature
- All 27 structures passed chemical validity checks
Critical revision (reversion risk analysis): The initial top candidate — the PEG prodrug (nominal green score 56.9) — was downgraded after reversion risk analysis. Prodrugs are designed to release the parent drug in vivo; once metabolized, the patient excretes persistent CBZ and its equally persistent metabolites. The true environmental green score of all prodrugs equals the parent's score (34.0). The corrected top candidates are the COOH-substituted variants (green score 48.8, strongest binding, novel, zero reversion risk).
Corrected headline metrics:
| Metric | Parent (CBZ) | Best Modification (corrected) | Improvement |
|---|---|---|---|
| Green Score | 34.0 | 48.8 (COOH at ring positions) | +44% |
| Reversion Risk | — | None (covalent C−COOH bond) | Permanent |
| Binding affinity | -2.79 kcal/mol | -3.12 kcal/mol | Maintained |
| Novelty | — | All 4 COOH isomers are novel | 100% new |
Figure 1. Carbamazepine (CBZ) — the parent compound. The dibenzoazepine core consists of two fused benzene rings (Ring A and Ring B) bridged by an azepine ring with a vinyl bridge (C=C) and a carboxamide group (-C(=O)NH₂) on the nitrogen.
Carbamazepine (5H-Dibenz[b,f]azepine-5-carboxamide)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
SMILES: NC(=O)N1c2ccccc2C=Cc2ccccc12
MW: 236.3 Da
LogP: 3.39
TPSA: 46.3 Ų
HBD/HBA: 1/1
H₂N
\
C=O ╱──╲ ╱──╲
╱ │ │──CH═CH──│ │
N │ │ │ │
╱ ╲ ╲──╱ ╲──╱
╱ ╲ Ring A Ring B
Ring A Ring B (positions (positions
unmodified) modified)
Environmental profile (baseline):
- Green score: 34.0 / 100 (poor)
- Biodegradation: Inherently biodegradable (t½ ≈ 138 days)
- BCF: 151 (moderate bioaccumulation)
- Fish LC50: 0.005 mg/L (high aquatic toxicity)
- Known metabolite: 10,11-epoxide (the primary Phase I metabolite)
Three independent strategies were applied to the parent compound. Red highlighting in molecular images indicates atoms/bonds that differ from the parent.
Figure 2. Overview of all three modification strategies applied to carbamazepine. Left: parent compound. Right: top 2 candidates from each strategy with structural differences highlighted in red.
Target: The urea bond (N−C(=O)−N) connecting the carboxamide to the dibenzoazepine nitrogen.
Rationale: Replacing or extending the urea bond with hydrolyzable linkers (ester, amide, ether) introduces enzymatic and abiotic cleavage sites, accelerating environmental degradation.
Reversion risk: MEDIUM — These compounds fragment into pieces that are not the parent CBZ molecule, but the resulting fragments have unverified environmental profiles (see Section 4).
Figure 3. All 8 biodegradable linker modifications. Parent compound (top-left, green label) with each modification showing the structural change highlighted in red.
PARENT: H₂N─C(=O)─N< (urea)
│
▼
┌────────────────────────────────────────────────────────────┐
│ Ester insertion: H₂N─C(=O)─O─C(=O)─N< │
│ Biuret extension: H₂N─C(=O)─NH─C(=O)─N< ★ Known │
│ Hydrazide: H₂N─NH─C(=O)─C(=O)─N< ★ G=52.3 │
│ Ether linkage: H₂N─CH₂─O─CH₂─C(=O)─N< │
│ (+ reversed variants of each) │
└────────────────────────────────────────────────────────────┘
| # | Modification | Green | Activity | Novel? | Reversion Risk |
|---|---|---|---|---|---|
| 1 | Ester linker (forward) | 48.6 | -3.06 | Novel | Medium |
| 2 | Hydroxamic acid variant | 43.4 | -2.93 | Novel | Medium |
| 3 | Biuret (double urea) | 44.8 | -3.04 | Known (EP Impurity C) | Medium |
| 4 | Hydrazide variant | 52.3 | -2.91 | Novel | Medium |
| 5 | Ether linker (forward) | 41.3 | -2.81 | Novel | Medium |
| 6 | Ester linker (reversed) | 43.4 | -2.93 | Novel | Medium |
| 7 | Hydrazide (reversed) | 52.3 | -2.91 | Novel | Medium |
| 8 | Ether linker (reversed) | 41.3 | -2.81 | Novel | Medium |
Finding: The biuret variant (#3) is already cataloged as Carbamazepine EP Impurity C (CAS 1219170-51-0), a known synthesis byproduct. The hydrazide variants (#4, #7) are novel and achieve the highest green scores in this strategy. However, their hydrolyzable bonds mean the green score applies to the intact form — hydrolysis fragments need separate environmental assessment.
Target: Aromatic C−H positions on Ring B of the dibenzoazepine.
Rationale: Adding polar substituents (−OH, −NH₂, −COOH, −SO₃H) reduces LogP, lowering bioaccumulation potential and improving aqueous solubility.
Reversion risk: NONE — These are covalent aromatic C−X bonds. They do not cleave under physiological or environmental conditions. The green score is real and permanent.
Figure 4. All 16 polar group modifications arranged in a grid. Four substituents (OH, NH₂, COOH, SO₃H) at four ring positions.
Ring B positions modified:
pos 1 pos 2 pos 3 pos 4
↓ ↓ ↓ ↓
╱──╲ │ ╱──╲ │ ╱──╲ │ ╱──╲
│ │ │ │ │ │ │ │ │ │ │
│ │──│─│ │──│─│ │──│─│ │
╲──╱ │ ╲──╱ │ ╲──╱ │ ╲──╱
* │ * │ * │ *
│ │ │
4 groups × 4 positions = 16 compounds
| Position | −OH | −NH₂ | −COOH | −SO₃H |
|---|---|---|---|---|
| Pos 1 | Known (CID 141058406) G:41.4 | Novel G:40.5 | Novel G:48.8 | Novel G:45.6 |
| Pos 2 | Known (CID 102174537) G:41.4 | Novel G:40.5 | Novel G:48.8 | Novel G:45.6 |
| Pos 3 | Known metabolite (3-OH-CBZ, CID 135290) G:41.4 | Known (CID 12783054) G:40.5 | Novel G:48.8 | Novel G:45.6 |
| Pos 4 | Known metabolite (2-OH-CBZ, CID 129274) G:41.4 | Novel G:40.5 | Novel G:48.8 | Novel G:45.6 |
Key findings:
- All 4 hydroxyl isomers are known in PubChem, but only 2-OH-CBZ and 3-OH-CBZ are well-characterized as metabolites detected in wastewater
- All COOH variants are novel across all positions — unexplored chemical space
- All SO₃H variants are novel — sulfonation of the CBZ scaffold has not been previously reported
- COOH addition gives the best green scores (48.8) among polar groups due to increased biodegradability
- All 16 compounds have zero reversion risk — permanent structural modifications
Target: The primary amine (−NH₂) of the carboxamide group.
Rationale: Masking the amine with cleavable protecting groups can improve environmental degradability of the intact prodrug while releasing active CBZ in vivo.
WARNING — Reversion risk: HIGH. Prodrugs are designed to release the parent drug after administration. The green score of the intact prodrug form is misleading for environmental assessment because the compound that actually reaches the environment (via patient excretion) is the parent CBZ and its persistent metabolites. See Section 4 for full analysis.
Figure 5. All 3 prodrug modifications. Left to right: parent, aldehyde masking, N-methylation, PEG conjugation.
PARENT: H₂N─C(=O)─N<
│
▼
┌───────────────────────────────────────────────────────────────────┐
│ Aldehyde: OHC─CH₂─NH─C(=O)─N< G: 40.8 Novel │
│ N-methyl: CH₃─NH─C(=O)─N< G: 33.2 Known │
│ PEG: HO─CH₂CH₂─O─CH₂CH₂─O─C(=O)─ G: 56.9 Novel │
│ CH₂─NH─C(=O)─N< ⚠ MISLEADING │
└───────────────────────────────────────────────────────────────────┘
| # | Prodrug Type | Nominal Green | True Green | Activity | Novel? | Reversion Risk |
|---|---|---|---|---|---|---|
| 25 | Aldehyde masking | 40.8 | 34.0 | -2.80 | Novel | HIGH |
| 26 | N-methylation | 33.2 | 34.0 | -2.68 | Known (CID 738915) | HIGH (CYP450 demethylation) |
| 27 | PEG conjugation | 34.0 | -3.10 | Novel | HIGH (esterase hydrolysis, t½ = 1-70h) |
Critical finding: The PEG prodrug's nominal green score of 56.9 only describes the intact PEG-CBZ conjugate. In the body, carboxylesterases cleave the PEG-ester bond (literature half-life: 1–70 hours), releasing unmodified CBZ. Since only 1–3% of CBZ is excreted unchanged (the rest becomes equally persistent metabolites like the 10,11-epoxide, 2-OH-CBZ, and 3-OH-CBZ), the true environmental green score of all prodrugs is 34.0 — identical to the parent.
This section evaluates whether each modification strategy produces compounds that remain modified when they reach the environment, or whether they revert to the persistent parent CBZ.
┌──────────────┐
│ PEG-CBZ │ Green Score: 56.9 (nominal)
│ Prodrug │ ← This is what the pipeline scored
└──────┬───────┘
│ Patient takes drug orally
▼
┌──────────────┐
│ Absorption │ PEG-ester bond intact in stomach
└──────┬───────┘
│ Reaches bloodstream
▼
┌──────────────┐
│ Esterase │ Carboxylesterase 1 & 2 (CES1/CES2)
│ Hydrolysis │ t½ = 1–70 hours (literature range)
└──────┬───────┘
│ Releases parent drug
▼
┌──────────────┐
│ CBZ │ ← Back to the parent compound
│ (unmodified) │ Green Score: 34.0 (the REAL score)
└──────┬───────┘
│ CYP3A4 / CYP2C8 metabolism
▼
┌────────────┼────────────┐
▼ ▼ ▼
┌──────────┐ ┌──────────┐ ┌──────────┐
│10,11- │ │ 2-OH-CBZ │ │ 3-OH-CBZ │
│Epoxide │ │ │ │ │
│(→Diol) │ │ Persist. │ │ Persist. │
└────┬─────┘ └────┬─────┘ └────┬─────┘
│ │ │
▼ ▼ ▼
┌─────────────────────────────────────┐
│ EXCRETED TO ENVIRONMENT │
│ All metabolites are persistent │
│ in wastewater treatment │
│ (Miao et al., 2005; Leclercq, 2009)│
└─────────────────────────────────────┘
Key pharmacokinetic facts (literature):
- Only 1–3% of CBZ is excreted unchanged; the rest is metabolized
- The primary metabolite (10,11-epoxide, 40–60% of dose) is also environmentally persistent
- 2-OH-CBZ and 3-OH-CBZ are detected in wastewater at 1.6–4.3 ug/kg (Miao et al., 2005)
- None of the major CBZ metabolites are readily biodegradable in wastewater treatment
| Strategy | # Compounds | Reversion Risk | Mechanism | True Green Score | Environmental Fate |
|---|---|---|---|---|---|
| Polar Group Addition | 16 | NONE | Covalent aromatic C−X bond; non-cleavable | As computed (40.5–48.8) | Modified compound reaches environment intact |
| Biodegradable Linker | 8 | MEDIUM | Hydrolyzable bonds (ester, amide, hydrazide) | Uncertain — fragments ≠ parent CBZ | Novel fragments; activity & fate unverified |
| Prodrug Design | 3 | HIGH | Designed for in vivo cleavage (esterase, CYP450) | 34.0 (= parent CBZ) | Parent CBZ + persistent metabolites |
After accounting for reversion risk, the true top candidates change significantly:
| Corrected Rank | SMILES | Strategy | Nominal Green | True Green | Activity | Novel? | Reversion Risk |
|---|---|---|---|---|---|---|---|
| 1 | NC(=O)N1c2ccccc2C=Cc2cccc(C(=O)O)c21 |
COOH at pos 2 | 48.8 | 48.8 | -3.12 | Novel | None |
| 2 | NC(=O)N1c2ccccc2C=Cc2ccc(C(=O)O)cc21 |
COOH at pos 3 | 48.8 | 48.8 | -3.12 | Novel | None |
| 3 | NC(=O)N1c2ccccc2C=Cc2c(C(=O)O)cccc21 |
COOH at pos 1 | 48.8 | 48.8 | -3.12 | Novel | None |
| 4 | NC(=O)N1c2ccccc2C=Cc2cc(C(=O)O)ccc21 |
COOH at pos 4 | 48.8 | 48.8 | -3.12 | Novel | None |
| 5 | NC(=O)N1c2ccccc2C=Cc2cccc(S(=O)(=O)O)c21 |
SO₃H at pos 2 | 45.6 | 45.6 | -3.25 | Novel | None |
Why COOH variants are the true top candidates:
- Zero reversion risk — covalent aromatic C−COOH bond is non-cleavable
- Strongest binding affinity of all 27 compounds (-3.12 kcal/mol)
- All four positional isomers are novel — no PubChem hits, no literature
- Green score of 48.8 is the highest reliable score (no reversion correction needed)
- Precedent: Oxcarbazepine (a 10-keto modification of CBZ) maintains anticonvulsant activity with a ring modification, supporting the viability of aromatic substitution on the CBZ scaffold
The hydrazide linker variants (nominal green score 52.3) occupy an intermediate position:
- Pro: The hydrolysis fragments are not the parent CBZ — the N−N hydrazide bond cleaves into two novel fragments
- Con: The environmental fate of these fragments is unknown and unverified
- Con: Whether the modified compound retains anticonvulsant activity (as an intact molecule) is unverified
- Status: MEDIUM risk — promising but requires experimental verification of both activity and fragment fate
This analysis aligns with the "benign by design" approach advocated by Kummerer (2007), which argues that the most effective environmental strategy for persistent pharmaceuticals is to modify the active drug molecule itself — not to create prodrugs or formulation-level changes.
Key principles:
- Modify the pharmacophore directly to introduce biodegradable features
- Retain pharmacological activity through structure-activity relationships
- Verify environmental fate of the modified drug and its metabolites
- Precedent: Kummerer's group demonstrated this approach with propranolol, creating derivatives with maintained beta-blocking activity but improved biodegradability
The polar group addition strategy (especially COOH) exemplifies this approach: the CBZ scaffold is directly modified with a permanent, non-cleavable substituent that improves environmental properties.
Figure 6. Activity vs Green Chemistry Score scatter plot. Note: prodrug compounds (if shown above the green threshold) have misleading green scores due to reversion risk.
| Category | Total | Known | Novel | % Novel |
|---|---|---|---|---|
| Biodegradable Linker | 8 | 1 | 7 | 88% |
| Polar Group (OH) | 4 | 4 | 0 | 0% |
| Polar Group (NH₂) | 4 | 1 | 3 | 75% |
| Polar Group (COOH) | 4 | 0 | 4 | 100% |
| Polar Group (SO₃H) | 4 | 0 | 4 | 100% |
| Prodrug | 3 | 1 | 2 | 67% |
| Total | 27 | 7 | 20 | 74% |
| Compound | PubChem CID | Status | Notes |
|---|---|---|---|
| 2-OH-CBZ | 129274 | Major metabolite | CAS 68011-66-5; detected in wastewater (1.9 ug/kg); extensively studied |
| 3-OH-CBZ | 135290 | Major metabolite | CAS 68011-67-6; detected in wastewater (4.3 ug/kg); very persistent in treatment |
| 1-OH-CBZ | 141058406 | Minor compound | CAS 68011-75-6; limited data |
| 4-OH-CBZ equiv. | 102174537 | Minor compound | CAS 68011-68-7; limited data |
| N-Carbamoyl CBZ | 71314504 | EP Impurity C | CAS 1219170-51-0; pharmacopeial standard |
| 2-Amino-CBZ | 12783054 | Registered | Limited data |
| N-Methyl CBZ | 738915 | Known | CHEMBL1594958; understudied |
Figure 7. Box plot comparing Green Chemistry Scores across the three modification strategies.
Figure 8. Property correlation heatmap showing relationships between activity, green score, biodegradation, BCF, and ecotoxicity scores.
Figure 9. Distribution plots for key scores across all 27 compounds.
Figure 10. Radar chart showing multi-dimensional property profiles of the top 5 candidates. Note: the original radar chart was generated before reversion correction — see corrected rankings in Section 4.3.
| Strategy | Count | Mean Nominal Green | Mean True Green | Mean Activity | Novel | Reversion Risk |
|---|---|---|---|---|---|---|
| Biodegradable Linker | 8 | 45.9 | Uncertain | -2.93 | 7 | Medium |
| Polar Group | 16 | 44.1 | 44.1 | -3.08 | 11 | None |
| Prodrug | 3 | 43.6 | 34.0 | -2.86 | 2 | High |
Polar group additions emerge as the most reliable strategy: their green scores are real (no reversion), they achieve the strongest mean binding affinity, and they contribute the most novel compounds. Prodrugs are downgraded to last place because their true environmental green score is identical to the parent compound.
| Rank | SMILES | Strategy | True Green | Activity | Novel? | Reversion Risk | Key Advantage |
|---|---|---|---|---|---|---|---|
| 1 | NC(=O)N1c2ccccc2C=Cc2cccc(C(=O)O)c21 |
COOH pos 2 | 48.8 | -3.12 | Novel | None | Best green + activity + no risk |
| 2 | NC(=O)N1c2ccccc2C=Cc2ccc(C(=O)O)cc21 |
COOH pos 3 | 48.8 | -3.12 | Novel | None | Same; different positional isomer |
| 3 | NC(=O)N1c2ccccc2C=Cc2c(C(=O)O)cccc21 |
COOH pos 1 | 48.8 | -3.12 | Novel | None | Same; different positional isomer |
| 4 | NC(=O)N1c2ccccc2C=Cc2cc(C(=O)O)ccc21 |
COOH pos 4 | 48.8 | -3.12 | Novel | None | Same; different positional isomer |
| 5 | NC(=O)N1c2ccccc2C=Cc2cccc(S(=O)(=O)O)c21 |
SO₃H pos 2 | 45.6 | -3.25 | Novel | None | Highest polarity; strong binding |
All top 5 candidates are novel, have zero reversion risk, and represent permanent structural modifications to the CBZ scaffold.
| Former Rank | SMILES | Strategy | Nominal Green | True Green | Issue |
|---|---|---|---|---|---|
O=C(CNC(=O)N1c2ccccc2C=Cc2ccccc21)OCCOCCO |
PEG prodrug | 34.0 | Reverts to CBZ (esterase hydrolysis) | ||
NNC(=O)C(=O)N1c2ccccc2C=Cc2ccccc21 |
Hydrazide | 52.3 | Uncertain | Fragment fate unverified | |
NC(=O)C(=O)NN1c2ccccc2C=Cc2ccccc21 |
Hydrazide (rev.) | 52.3 | Uncertain | Fragment fate unverified |
| Check | Result |
|---|---|
| SMILES validity | 27/27 pass |
| Property ranges | 0 violations |
| Statistical outliers | 3 detected (not blocking) |
| Reversion risk assessment | Completed — 3 strategies classified |
| Overall | PASS |
-
Synthesize COOH-substituted carbamazepine variants — all four ring-B positional isomers. These are the optimal candidates because:
- Novel (100% — no prior literature)
- Zero reversion risk (covalent aromatic C−COOH)
- Highest reliable green score (48.8)
- Strongest binding affinity of all 27 compounds (-3.12 kcal/mol)
- Precedent: Oxcarbazepine demonstrates that CBZ scaffold modifications maintain anticonvulsant activity
-
Test anticonvulsant activity of COOH-CBZ variants in vitro (Nav1.7 binding assay) and in vivo (MES test in rodents) to confirm the computational binding predictions.
-
Measure environmental biodegradation of COOH-CBZ using OECD 301 (ready biodegradability) and OECD 302 (inherent biodegradability) standard tests.
-
SO₃H-substituted variants as secondary candidates — novel, zero reversion risk, green score 45.6, but lower than COOH variants.
-
Hydrazide linker compounds — highest nominal green scores (52.3) but require experimental verification that: (a) the intact hydrazide retains anticonvulsant activity, and (b) the hydrolysis fragments are themselves biodegradable.
- Prodrug approaches should be deprioritized for environmental green chemistry. While prodrugs remain valuable for pharmacokinetic optimization (improving oral bioavailability, reducing GI side effects, etc.), they do not solve the environmental persistence problem because the parent drug is released in the body and its persistent metabolites are excreted.
- Adopt the "benign by design" approach (Kummerer, 2007) for future pharmaceutical green chemistry: modify the active drug molecule itself rather than creating cleavable prodrug forms. The COOH-CBZ variants exemplify this principle.
- Data collection: ChEMBL API (174 bioactivity records, target CHEMBL2094253) + PubChem similarity search (CID 2554, threshold 80%)
- Molecular descriptors: RDKit (MW, LogP, TPSA, HBD, HBA, rotatable bonds, aromatic rings)
- Binding affinity: Descriptor-based proxy scoring with MMFF94 3D optimization; protein target: Nav1.7 (PDB: 6J8E)
- Environmental fate: BIOWIN-like biodegradation, Veith BCF regression, Konemann narcosis-baseline ecotoxicity, SMARTS-based metabolite prediction
- Green score: Composite 0-100 metric: 40% biodegradation + 30% BCF (inverted) + 30% ecotoxicity
- Novelty: PubChem PUG-REST exact structure search + web literature review
- Reversion risk: Bond-type analysis (covalent aromatic vs. hydrolyzable vs. enzymatically cleavable) + pharmacokinetic literature review
- Visualization: RDKit MolDraw2D with MCS-based difference highlighting
- PubChem Compound Database. https://pubchem.ncbi.nlm.nih.gov/
- Miao, X.S. et al. (2005) "Carbamazepine and Its Metabolites in Wastewater and in Biosolids." Environ. Sci. Technol. DOI:10.1021/es050261e
- Leclercq, M. et al. (2009) "Presence and fate of carbamazepine and its metabolites at wastewater treatment plants." Arch. Environ. Contam. Toxicol.
- European Pharmacopeia — Carbamazepine monograph (impurity specifications)
- BOC Sciences — Carbamazepine Impurities Reference Standards
- Kummerer, K. (2007) "Sustainable from the very beginning: rational design of molecules by life cycle engineering as an important approach for green pharmacy and green chemistry." Green Chemistry 9:899–907. DOI:10.1039/B618298B
- Kummerer, K. & Hempel, M. (2010) Green and Sustainable Pharmacy. Springer-Verlag Berlin Heidelberg.
- Lienert, J. et al. (2007) "Reducing micropollutants with source control: substance flow analysis of 212 pharmaceuticals in faeces and urine." Water Sci. Technol. 56(5):87–96.
- Beaumont, K. et al. (2003) "Design of ester prodrugs to enhance oral absorption of poorly permeable compounds." Pharm. Res. 20:1589–96.
- Imai, T. & Ohura, K. (2010) "The role of intestinal carboxylesterase in the oral absorption of prodrugs." Curr. Drug Metab. 11(9):793–805.
- Testa, B. & Mayer, J.M. (2003) Hydrolysis in Drug and Prodrug Metabolism. Wiley-VCH.
- Friberg, L.E. et al. (2009) "Mechanistic models for myelosuppression." Invest. New Drugs 21:183–194.
| File | Description |
|---|---|
reports/parent_carbamazepine.png |
Figure 1 — Parent structure |
reports/modification_pathways.png |
Figure 2 — Overview pathway diagram |
reports/pathway_biodegradable_linker.png |
Figure 3 — Linker modifications grid |
reports/pathway_polar_group.png |
Figure 4 — Polar group modifications grid |
reports/pathway_prodrug.png |
Figure 5 — Prodrug modifications grid |
reports/activity_vs_green.png |
Figure 6 — Activity vs Green scatter |
reports/modification_comparison.png |
Figure 7 — Strategy comparison box plot |
reports/correlation_heatmap.png |
Figure 8 — Property correlations |
reports/distributions.png |
Figure 9 — Score distributions |
reports/radar_top_candidates.png |
Figure 10 — Top candidate profiles |
reports/pathway_diagram.txt |
Text-based pathway diagram (ASCII) |
reports/novelty_assessment.md |
Detailed novelty assessment with CIDs |
data/results/final_results.csv |
Full results (27 compounds, 9 columns) |
data/results/optimized_compounds.csv |
Ranked by multi-objective score |