PepALD: Macrocyclic Peptide Generation via Autoregressive Latent Diffusion

2026-06-12

AI Summary

This work addresses the challenge of simultaneously optimizing non-canonical monomers, macrocyclic topology, membrane permeability, and target affinity in macrocyclic peptide design by introducing the first autoregressive latent diffusion model. Built upon HELM-based chemical embeddings, the method generates sequences residue-by-residue in a chemically informed latent space while concurrently predicting side chain–aware cyclization patterns. It further integrates a Winner-Protected preference optimization mechanism to align sequence generation with affinity-based reward signals. Experimental results demonstrate that the model significantly outperforms existing peptide generation baselines in both generative quality and target affinity optimization, achieving, for the first time, structure-aware, end-to-end de novo design of macrocyclic peptides.

Abstract

Macrocyclic peptides are promising therapeutic candidates for intracellular targets, but their design requires simultaneous control over non-natural monomer chemistry, ring topology, membrane permeability, and target binding. Existing SMILES- or HELM-string generative models either operate in long atom-level sequence spaces or treat monomers as symbolic tokens with limited chemical grounding. We introduce PepALD, an Autoregressive Latent Diffusion (ALD) foundation model for \textit{de novo} macrocyclic peptide generation. The model represents HELM monomers with structured chemical embeddings, generates each residue through context-conditioned diffusion in chemically informed latent space, predicts R-group-aware ring closures during autoregressive generation, and aligns the denoiser to affinity rewards using winner-protected diffusion-adapted preference optimization. In silico experiments demonstrate PepALD's generation quality and reward-optimization performance against representative peptide generation baselines.

Problem

• macrocyclic peptides
• de novo design
• membrane permeability
• target binding
• non-natural monomers

Innovation

• Autoregressive Latent Diffusion
• Macrocyclic Peptide Generation
• Chemical Embeddings
• Ring Closure Prediction
• Preference Optimization