arXiv：cs.LG（机器学习，全量分类）· Aaron L. Feller, Andrew D. Ellington, Claus O. Wilke·· 13 小时前AI 评分34
StabilityArc：将蛋白质序列嵌入解码为可泛化稳定性图谱
StabilityArc: Decoding Protein Sequence Embeddings into Generalizable Stability Landscapes
AI 导读
StabilityArc 通过共享 RoPE transformer 将冻结的 ESMC-600M 残基表示映射为 Lx20 替换效应矩阵，在 66 项严格留一蛋白质评估、覆盖 134,794 个 ProteinGym 变体上达到 0.7134 Spearman 相关性，超过最强零样本基线 ProSST-2048（0.6526）0.0608。
正文
Abstract:Every protein has a unique stability landscape, but the physical consequences of mutation are governed by recurring biochemical constraints. We test whether a shared decoder, trained on measurements from diverse proteins, can interpret these constraints in an unseen target, enabling cross-protein transfer for initial experimental round prescreening. We present StabilityArc , which maps frozen ESMC-600M residue representations through a shared RoPE transformer to an Lx20 matrix of substitution effects; a symmetric, contact-aware residual aids in predicting epistasis in simultaneous substitutions. In 66 strict leave-one-protein-out evaluations covering 134,794 ProteinGym variants, StabilityArc achieves 0.7134 Spearman correlation, exceeding the strongest zero-shot baseline, ProSST-2048 (0.6526), by 0.0608. We further explore the utility of this method by providing the score as a prior for Kermut, achieving Spearman correlation of 0.8280 across three supervised split schemes, improving on Kermut's reported 0.8167.
| Comments: | Accepted to Representations for the Physical Sciences Workshop @ NeurIPS 2026; 9 pages, 1 figure, 2 tables |
| Subjects: | Biomolecules (q-bio.BM); Machine Learning (cs.LG) |
| MSC classes: | 92-08 |
| ACM classes: | I.2; J.2; J.3 |
| Cite as: | arXiv:2610.00742 [q-bio.BM] |
| (or arXiv:2610.00742v1 [q-bio.BM] for this version) | |
| https://doi.org/10.48550/arXiv.2610.00742 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Aaron Feller [view email]
[v1]
Wed, 30 Sep 2026 21:34:32 UTC (477 KB)
来源：arXiv：cs.LG（机器学习，全量分类） · arxiv.org