CupOFLATTE: Coupled Objective-Guided Discrete Flows via Linker Assembly for Targeted PROTAC Engineering

Ziang Li, Ruoxi Zhang, Pranam Chatterjee

Keywords: discrete flow matching, PROTAC linker design, constrained generation, molecular graph generation, targeted protein degradation

TL;DR: CupOFLATTE generates PROTAC linkers as constrained, objective-guided discrete flows over molecular graphs, outperforming existing linker-design baselines on validity, predicted activity, and novelty.

Abstract: Proteolysis-targeting chimeras (PROTACs) are heterobifunctional molecules whose function emerges from three coupled components: a warhead that binds a protein of interest, a linker, and an E3 ligand that recruits the cellular degradation machinery. Most generative approaches design the linker in isolation and assemble the ternary molecule only after generation, ignoring cross-component dependencies and often producing PROTACs that fall outside the empirical degrader distribution. We introduce **C**o**up**led **O**bjective-Guided Discrete **F**lows via **L**inker **A**ssembly for **T**argeted pro**T**ac **E**ngineering (**CupOFLATTE**), a discrete generative framework that restricts generative trajectories to a chemically feasible conjugate manifold and biases local transitions using soft objectives derived from PROTAC-DB chemistry, keeping warhead, linker, and E3 ligand components compatible throughout generation. Across assembly validity, predicted degradation activity (DegradeMaster), and PROTAC-like physicochemical property ranges, CupOFLATTE outperforms staged baselines including LinkerNet, DiffLinker, and DiffPROTACs, demonstrating that explicit coupling and constraint enforcement are sufficient to recover functional degraders in settings where function emerges only at the level of the assembled ternary system.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 50