arXiv：cs.LG· Takashi Fujiwara, Hikaru Shindo, Kaushalya Madhawa, Jun Jin Choong, Shuan Chen, Yuna Oikawa, Yiming Zhang, Gyubok Lee, Keisuke Ozawa·· 2 天前AI 评分39
Pepti-drift：无需推理时引导的可扩展安全活性肽生成框架
Pepti-drift: Scalable Safe-Active Peptide Generation Without Inference-Time Guidance
AI 导读
研究团队提出 Pepti-drift 一步生成框架，通过单次潜在精炼加并行解码实现安全活性肽生成，无需推理时引导。在 88 个留出靶点上，其预测 Safe-Active 产率达 18.37%，生成成本比多属性引导基线低 468 倍。团队同时发布 BindSafe-PepBench 基准，联合评估靶点结合与四项安全指标，并揭示肽长度是结合-安全联合评估的主要混杂因素。
正文
Abstract:Therapeutic peptides are a promising drug modality, but their generation must satisfy multiple therapeutic constraints. We introduce BindSafe-PepBench, a fixed-budget benchmark that jointly evaluates target binding and four major safety metrics on the same generated candidates. We reveal that peptide length is a major confounder of joint binding-safety evaluation: longer peptides tend toward stronger predicted binding but less favorable predicted safety. This creates an apparent trade-off and can bias comparisons among models with different output-length distributions. We report absolute Safe-Active yield and exact-length-matched gains to distinguish generative improvements from output-length effects. High Safe-Active yield remains challenging, while the strongest multi-property methods rely on costly inference-time guidance. We therefore introduce Pepti-drift, a one-step generation framework that incorporates attraction toward target-specific binders and repulsion from liability-associated regions, requiring a single latent refinement followed by parallel decoding without inference-time guidance. Across 88 held-out targets, Pepti-drift achieves an 18.37% predicted Safe-Active yield while retaining positive exact-length-matched gains. The resulting gains are competitive with multi-property-guided baselines while requiring 468 times lower generation cost, enabling scalable and fair high-throughput peptide design.
| Comments: | preprint |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.27824 [cs.LG] |
| (or arXiv:2606.27824v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2606.27824 arXiv-issued DOI via DataCite |
Submission history
From: Hikaru Shindo [view email]
[v1]
Fri, 26 Jun 2026 08:04:11 UTC (3,523 KB)
[v2]
Mon, 29 Jun 2026 01:23:29 UTC (3,523 KB)
[v3]
Thu, 1 Oct 2026 08:42:56 UTC (6,377 KB)
来源：arXiv：cs.LG · arxiv.org