Freeze, Diffuse, Decode: Geometry-Aware Adaptation of Pretrained Transformer Embeddings for Antimicrobial Peptide Design
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912735581700096 |
|---|---|
| author | Gawade, Pankhil Izdebski, Adam Lizotte, Myriam Moon, Kevin R. Rhodes, Jake S. Wolf, Guy Szczurek, Ewa |
| author_facet | Gawade, Pankhil Izdebski, Adam Lizotte, Myriam Moon, Kevin R. Rhodes, Jake S. Wolf, Guy Szczurek, Ewa |
| contents | Pretrained transformers provide rich, general-purpose embeddings, which are transferred to downstream tasks. However, current transfer strategies: fine-tuning and probing, either distort the pretrained geometric structure of the embeddings or lack sufficient expressivity to capture task-relevant signals. These issues become even more pronounced when supervised data are scarce. Here, we introduce Freeze, Diffuse, Decode (FDD), a novel diffusion-based framework that adapts pre-trained embeddings to downstream tasks while preserving their underlying geometric structure. FDD propagates supervised signal along the intrinsic manifold of frozen embeddings, enabling a geometry-aware adaptation of the embedding space. Applied to antimicrobial peptide design, FDD yields low-dimensional, predictive, and interpretable representations that support property prediction, retrieval, and latent-space interpolation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_23120 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Freeze, Diffuse, Decode: Geometry-Aware Adaptation of Pretrained Transformer Embeddings for Antimicrobial Peptide Design Gawade, Pankhil Izdebski, Adam Lizotte, Myriam Moon, Kevin R. Rhodes, Jake S. Wolf, Guy Szczurek, Ewa Machine Learning 68T01 Pretrained transformers provide rich, general-purpose embeddings, which are transferred to downstream tasks. However, current transfer strategies: fine-tuning and probing, either distort the pretrained geometric structure of the embeddings or lack sufficient expressivity to capture task-relevant signals. These issues become even more pronounced when supervised data are scarce. Here, we introduce Freeze, Diffuse, Decode (FDD), a novel diffusion-based framework that adapts pre-trained embeddings to downstream tasks while preserving their underlying geometric structure. FDD propagates supervised signal along the intrinsic manifold of frozen embeddings, enabling a geometry-aware adaptation of the embedding space. Applied to antimicrobial peptide design, FDD yields low-dimensional, predictive, and interpretable representations that support property prediction, retrieval, and latent-space interpolation. |
| title | Freeze, Diffuse, Decode: Geometry-Aware Adaptation of Pretrained Transformer Embeddings for Antimicrobial Peptide Design |
| topic | Machine Learning 68T01 |
| url | https://arxiv.org/abs/2511.23120 |