Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
[Preprint]. 2025 Jun 2:arXiv:2412.17780v4.

PepTune: De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusion

Sophia Tang  1 , Yinuo Zhang  2 , Pranam Chatterjee  1   3
Affiliations

PepTune: De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusion

Sophia Tang et al. ArXiv. .

Abstract

We present PepTune, a multi-objective discrete diffusion model for simultaneous generation and optimization of therapeutic peptide SMILES. Built on the Masked Discrete Language Model (MDLM) framework, PepTune ensures valid peptide structures with a novel bond-dependent masking schedule and invalid loss function. To guide the diffusion process, we introduce Monte Carlo Tree Guidance (MCTG), an inference-time multi-objective guidance algorithm that balances exploration and exploitation to iteratively refine Pareto-optimal sequences. MCTG integrates classifier-based rewards with search-tree expansion, overcoming gradient estimation challenges and data sparsity. Using PepTune, we generate diverse, chemically-modified peptides simultaneously optimized for multiple therapeutic properties, including target binding affinity, membrane permeability, solubility, hemolysis, and non-fouling for various disease-relevant targets. In total, our results demonstrate that MCTG for masked discrete diffusion is a powerful and modular approach for multi-objective sequence design in discrete state spaces.

PubMed Disclaimer

Conflict of interest statement

Competing Interests P.C. is a co-founder of Gameto, Inc. and UbiquiTx, Inc. and advises companies involved in peptide therapeutics development. P.C., S.T., and Y.Z. have and are currently filing patent applications related to this work. P.C.’s interests are reviewed and managed by Duke University in accordance with their conflict-of-interest policies.

Figures

Figure 1.
Figure 1.. PepMDLM.
PepMDLM is a discrete masked diffusion model for unconditional de novo generation of peptide SMILES representations.
Figure 2.
Figure 2.. PepTune.
PepTune is a multi-objective discrete diffusion model with Monte Carlo Tree Guidance (MCTG). The full algorithm is detailed in Algorithm 3.
Figure 3.
Figure 3.. Comparison of docked PepTune-generated peptides to existing GLP-1R agonists.
(A, B) Docking images of semaglutide (score: −5.7 kcal/mol) and liraglutide (score: −5.1 kcal/mol) binding to GLP-1R. (C) Full view of the positive control GLP-1R agonists and the PepTune-generated binders on GLP-1R. (D, E) Docking images of binder 1 (score: −7.4 kcal/mol) and 2 (score: −7.0 kcal/mol) were generated using PepTune, conditioned on predicted affinity to GLP-1R, solubility, hemolysis, and non-fouling. Shared polar contacts between binder 1 and either controls are highlighted in pink, shared polar contacts between binder 2 and either controls are highlighted in green, and the shared contacts across both binders are highlighted in purple.
Figure 4.
Figure 4.. PepTune-generated peptides to TfR and GLAST.
Full protein binding location and close-up binding position for (A) dual binder 1, (B) dual binder 6, and (C) dual binder 8 with TfR (left) and GLAST (right). Polar contacts within 3.5 Å are highlighted.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Alfaris N., Waldrop S., Johnson V., Boaventura B., Kendrick K., and Stanford F. C. Glp-1 single, dual, and triple receptor agonists for treating type 2 diabetes and obesity: a narrative review. eClinicalMedicine, 75: 102782, 2024. - PMC - PubMed
    1. Austin J., Johnson D. D., Ho J., Tarlow D., and Berg R. v. d. Structured denoising diffusion models in discrete state-spaces. Advances in Neural Information Processing Systems, 2021.
    1. Bhat S., Palepu K., Hong L., Mao J., Ye T., Iyer R., Zhao L., Chen T., Vincoff S., and Watson R. e. a. De novo design of peptide binders to conformationally diverse targets with contrastive language modeling. Science Advances, 11(4), 2025. - PMC - PubMed
    1. Bi Y., Liu L., Lu Y., Sun T., Shen C., Chen X., Chen Q., An S., He X., and Ruan C. e. a. T7 peptide-functionalized peg-plga micelles loaded with carmustine for targeting therapy of glioma. ACS Applied Materials & Interfaces, 8(41):27465–27473, 2016. - PubMed
    1. Brenner M., Johnson A. B., Boespflug-Tanguy O., Rodriguez D., Goldman J. E., and Messing A. Mutations in gfap, encoding glial fibrillary acidic protein, are associated with alexander disease. Nature Genetics, 27(1): 117–120, 2001. - PubMed

Publication types

LinkOut - more resources