PepRePs: Peptide-Retargeted Phosphatases via Generative Language Models

Lauren Hong, Tai-Chen Ho, Yi-Shiuan Tseng, Sophia Vincoff, Tong Chen, Pohan chen, Pranam Chatterjee

Confirmation: I have read and agree with the workshop's policy on behalf of myself and my co-authors.

Track: tiny / short paper (2-4 pages excluding references; extended abstract format)

Keywords: Cell state modeling, protein language models, targeted dephosphorylation, proteome editing

TL;DR: We use language model–designed peptides to selectively dephosphorylate proteins, enabling programmable control over phosphorylation-driven cellular states.

Abstract: Phosphorylation encodes a fast, reversible layer of regulation that directly governs protein activity, signaling flow, and cell state transitions. While kinases and phosphatases collectively shape these landscapes, existing tools lack the ability to selectively rewrite phosphorylation states of individual, often disordered, proteins inside cells. Here, we introduce **Pep**tide-**Re**targeted **P**hosphatases (**PepRePs**), a genetically encodable proteome-editing platform that uses protein language model-derived peptide binders to localize phosphatase activity to user-specified targets. PepRePs enable targeted dephosphorylation of endogenous and exogenous proteins, including Rab8a and tau, reducing site-specific phosphorylation and producing downstream functional effects linked to altered cellular behavior. By coupling programmable peptide binders with modular phosphatase domains, PepRePs provide a mechanism-aware way to perturb signaling networks at the level of post-translational control rather than gene expression. This work positions targeted dephosphorylation as a scalable strategy for probing and reprogramming phosphorylation-driven cell states, with implications for both systems biology and therapeutic intervention in diseases dominated by signaling dysregulation.

Anonymization: This submission has been anonymized for double-blind review via the removal of identifying information such as names, affiliations, and identifying URLs.

Submission Number: 6