![[2510.01988] PepCompass: Navigating peptide embedding spaces using Riemannian Geometry](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2510.01988] PepCompass: Navigating peptide embedding spaces using Riemannian Geometry
Summary
PepCompass introduces a geometry-aware framework for exploring peptide spaces, enhancing antimicrobial peptide discovery through advanced Riemannian geometry techniques.
Why It Matters
This research addresses the challenges in antimicrobial peptide discovery by proposing a novel approach that leverages Riemannian geometry, potentially leading to more effective treatments against resistant bacterial strains. The findings could significantly impact drug design and development in the field of biotechnology.
Key Takeaways
- PepCompass utilizes Riemannian geometry to improve peptide exploration.
- The framework introduces two local exploration methods for efficient optimization.
- In-vitro validation shows the discovery of novel antimicrobial peptides.
- Geometry-informed approaches can enhance the design of effective drugs.
- The study highlights the importance of adapting models to the intrinsic dimensionality of peptide data.
Computer Science > Machine Learning arXiv:2510.01988 (cs) [Submitted on 2 Oct 2025 (v1), last revised 25 Feb 2026 (this version, v5)] Title:PepCompass: Navigating peptide embedding spaces using Riemannian Geometry Authors:Marcin Możejko, Adam Bielecki, Jurand Prądzyński, Marcin Traskowski, Antoni Janowski, Hyun-Su Lee, Marcelo Der Torossian Torres, Michał Kmicikiewicz, Paulina Szymczak, Karol Jurasz, Michał Kucharczyk, Cesar de la Fuente-Nunez, Ewa Szczurek View a PDF of the paper titled PepCompass: Navigating peptide embedding spaces using Riemannian Geometry, by Marcin Mo\.zejko and 12 other authors View PDF HTML (experimental) Abstract:Antimicrobial peptide discovery is challenged by the astronomical size of peptide space and the relative scarcity of active peptides. Generative models provide continuous latent "maps" of peptide space, but conventionally ignore decoder-induced geometry and rely on flat Euclidean metrics, rendering exploration and optimization distorted and inefficient. Prior manifold-based remedies assume fixed intrinsic dimensionality, which critically fails in practice for peptide data. Here, we introduce PepCompass, a geometry-aware framework for peptide exploration and optimization. At its core, we define a Union of $\kappa$-Stable Riemannian Manifolds $\mathbb{M}^{\kappa}$, a family of decoder-induced manifolds that captures local geometry while ensuring computational stability. We propose two local exploration methods: Second-Order Riemannian Brown...
