Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866914304970719232 |
|---|---|
| author | Kim, Hyeonah Kim, Minsu Roget, Celine Biton, Dionessa Vaillancourt, Louis Brun, Yves V. Bengio, Yoshua Hernandez-Garcia, Alex |
| author_facet | Kim, Hyeonah Kim, Minsu Roget, Celine Biton, Dionessa Vaillancourt, Louis Brun, Yves V. Bengio, Yoshua Hernandez-Garcia, Alex |
| contents | The application of generative models for experimental drug discovery campaigns is severely limited by the difficulty of designing molecules de novo that can be synthesized in practice. Previous works have leveraged Generative Flow Networks (GFlowNets) to impose hard synthesizability constraints through the design of state and action spaces based on predefined reaction templates and building blocks. Despite the promising prospects of this approach, it currently lacks flexibility and scalability. As an alternative, we propose S3-GFN, which generates synthesizable SMILES molecules via simple soft regularization of a sequence-based GFlowNet. Our approach leverages rich molecular priors learned from large-scale SMILES corpora to steer molecular generation towards high-reward, synthesizable chemical spaces. The model induces constraints through off-policy replay training with a contrastive learning signal based on separate buffers of synthesizable and unsynthesizable samples. Our experiments show that S3-GFN learns to generate synthesizable molecules ($\geq 95\%$) with higher rewards in diverse tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_04119 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors Kim, Hyeonah Kim, Minsu Roget, Celine Biton, Dionessa Vaillancourt, Louis Brun, Yves V. Bengio, Yoshua Hernandez-Garcia, Alex Machine Learning Quantitative Methods The application of generative models for experimental drug discovery campaigns is severely limited by the difficulty of designing molecules de novo that can be synthesized in practice. Previous works have leveraged Generative Flow Networks (GFlowNets) to impose hard synthesizability constraints through the design of state and action spaces based on predefined reaction templates and building blocks. Despite the promising prospects of this approach, it currently lacks flexibility and scalability. As an alternative, we propose S3-GFN, which generates synthesizable SMILES molecules via simple soft regularization of a sequence-based GFlowNet. Our approach leverages rich molecular priors learned from large-scale SMILES corpora to steer molecular generation towards high-reward, synthesizable chemical spaces. The model induces constraints through off-policy replay training with a contrastive learning signal based on separate buffers of synthesizable and unsynthesizable samples. Our experiments show that S3-GFN learns to generate synthesizable molecules ($\geq 95\%$) with higher rewards in diverse tasks. |
| title | Synthesizable Molecular Generation via Soft-constrained GFlowNets with Rich Chemical Priors |
| topic | Machine Learning Quantitative Methods |
| url | https://arxiv.org/abs/2602.04119 |