Update README.md
Browse files
README.md
CHANGED
|
@@ -13,9 +13,6 @@ license: mit
|
|
| 13 |
- **Languages:** SMARTS & SMILES (chemical substructure representation)
|
| 14 |
- **Pipeline Tag:** text2text-generation for MMP transformation
|
| 15 |
- **Library:** Transformers, PyTorch
|
| 16 |
-
|
| 17 |
-
---
|
| 18 |
-
|
| 19 |
# 2. Intended Use
|
| 20 |
|
| 21 |
- **Direct Use:**
|
|
@@ -43,18 +40,12 @@ license: mit
|
|
| 43 |
- **MMP-C2V:**
|
| 44 |
- Integration into analog enumeration pipelines
|
| 45 |
- Integration into high-throughput virtual screening pipelines
|
| 46 |
-
|
| 47 |
-
---
|
| 48 |
-
|
| 49 |
# 3. Bias, Risks, and Limitations
|
| 50 |
|
| 51 |
- **Known Limitations:** The models rely on the availability and coverage of large historical transformation datasets, and its performance may vary in underrepresented chemical domains.
|
| 52 |
- **Biases:** Inherits biases from ChEMBL-derived medicinal chemistry literature.
|
| 53 |
- **Risk Areas:** Our framework is intended for research use and does not introduce specific ethical concerns.
|
| 54 |
- **Recommendations:** None
|
| 55 |
-
|
| 56 |
-
---
|
| 57 |
-
|
| 58 |
# 4. Training Details
|
| 59 |
|
| 60 |
- **Training Data:** Raw data is downloaded from ChEMBL database and available at [https://chembl.gitbook.io/chembl-interface-documentation/downloads](https://chembl.gitbook.io/chembl-interface-documentation/downloads).
|
|
@@ -72,9 +63,6 @@ license: mit
|
|
| 72 |
- Hardware:
|
| 73 |
- MMPT-FM: 4 × NVIDIA A6000 GPUs
|
| 74 |
- MMP variants: 4 × NVIDIA H100 GPUs
|
| 75 |
-
|
| 76 |
-
---
|
| 77 |
-
|
| 78 |
# 5. Evaluation
|
| 79 |
|
| 80 |
- **Metrics:**
|
|
@@ -86,16 +74,10 @@ license: mit
|
|
| 86 |
- Within-patent analog generation (PMV17)
|
| 87 |
- Cross-patent analog generation (PMV17 → PMV21)
|
| 88 |
- **Testing Data:** Patent-derived datasets from PMV Pharmaceuticals (2017, 2021)
|
| 89 |
-
|
| 90 |
-
---
|
| 91 |
-
|
| 92 |
# 6. Usage
|
| 93 |
|
| 94 |
- **Sample Inference Code:** Described conceptually in the publications; code corresponds to variable-to-variable generation with beam search can be found at our GitHub repository: [https://github.com/MSDLLCpapers/MMPTTransformer](https://github.com/MSDLLCpapers/MMPTTransformer).
|
| 95 |
- **GitHub Links:** [https://github.com/MSDLLCpapers/MMPTTransformer](https://github.com/MSDLLCpapers/MMPTTransformer)
|
| 96 |
-
|
| 97 |
-
---
|
| 98 |
-
|
| 99 |
# 7. Citation
|
| 100 |
|
| 101 |
**BibTeX:**
|
|
|
|
| 13 |
- **Languages:** SMARTS & SMILES (chemical substructure representation)
|
| 14 |
- **Pipeline Tag:** text2text-generation for MMP transformation
|
| 15 |
- **Library:** Transformers, PyTorch
|
|
|
|
|
|
|
|
|
|
| 16 |
# 2. Intended Use
|
| 17 |
|
| 18 |
- **Direct Use:**
|
|
|
|
| 40 |
- **MMP-C2V:**
|
| 41 |
- Integration into analog enumeration pipelines
|
| 42 |
- Integration into high-throughput virtual screening pipelines
|
|
|
|
|
|
|
|
|
|
| 43 |
# 3. Bias, Risks, and Limitations
|
| 44 |
|
| 45 |
- **Known Limitations:** The models rely on the availability and coverage of large historical transformation datasets, and its performance may vary in underrepresented chemical domains.
|
| 46 |
- **Biases:** Inherits biases from ChEMBL-derived medicinal chemistry literature.
|
| 47 |
- **Risk Areas:** Our framework is intended for research use and does not introduce specific ethical concerns.
|
| 48 |
- **Recommendations:** None
|
|
|
|
|
|
|
|
|
|
| 49 |
# 4. Training Details
|
| 50 |
|
| 51 |
- **Training Data:** Raw data is downloaded from ChEMBL database and available at [https://chembl.gitbook.io/chembl-interface-documentation/downloads](https://chembl.gitbook.io/chembl-interface-documentation/downloads).
|
|
|
|
| 63 |
- Hardware:
|
| 64 |
- MMPT-FM: 4 × NVIDIA A6000 GPUs
|
| 65 |
- MMP variants: 4 × NVIDIA H100 GPUs
|
|
|
|
|
|
|
|
|
|
| 66 |
# 5. Evaluation
|
| 67 |
|
| 68 |
- **Metrics:**
|
|
|
|
| 74 |
- Within-patent analog generation (PMV17)
|
| 75 |
- Cross-patent analog generation (PMV17 → PMV21)
|
| 76 |
- **Testing Data:** Patent-derived datasets from PMV Pharmaceuticals (2017, 2021)
|
|
|
|
|
|
|
|
|
|
| 77 |
# 6. Usage
|
| 78 |
|
| 79 |
- **Sample Inference Code:** Described conceptually in the publications; code corresponds to variable-to-variable generation with beam search can be found at our GitHub repository: [https://github.com/MSDLLCpapers/MMPTTransformer](https://github.com/MSDLLCpapers/MMPTTransformer).
|
| 80 |
- **GitHub Links:** [https://github.com/MSDLLCpapers/MMPTTransformer](https://github.com/MSDLLCpapers/MMPTTransformer)
|
|
|
|
|
|
|
|
|
|
| 81 |
# 7. Citation
|
| 82 |
|
| 83 |
**BibTeX:**
|