Vilya-1 Foundation Model Predicts and Designs Macrocycle Structures.

Vilya Research, :, Pascal Sturmfels, Milad Salem, Naozumi Hiranuma, Stephen Rettie, Xiaoliang Pan, Benjamin D. Sellers, Adam P. Moyer, Patrick J. Salveson, Ivan Anishchanka· July 14, 2026 View original

Key takeaways

  • Vilya-1 is a new all-atom foundation model for macrocycle structure prediction.
  • It significantly improves geometric accuracy and predicts developability properties.
  • The model supports generative design of novel macrocycles with tailored profiles.
  • Vilya-1 accelerates the development of next-generation macrocycle therapeutics.

Who benefits

PharmaceuticalsBiotechnologyDrug DiscoveryHealthcare

Summary

Vilya-1 is a deep learning foundation model designed for predicting macrocycle structures and key developability properties like membrane permeability. It operates on an all-atom representation, trained on diverse structural datasets, and significantly improves geometric accuracy over existing methods while enabling generative design of novel macrocycles.

Researchers have developed Vilya-1, an all-atom deep learning foundation model aimed at revolutionizing macrocycle structure prediction and design. Macrocyclic peptides are increasingly vital in therapeutics, but current computational methods struggle with generalization across varied chemical spaces. Vilya-1 addresses this by uniformly representing all atoms and training on heterogeneous structural datasets, encompassing diverse topologies and chemical classes. The model excels at sampling biologically relevant conformations and predicting crucial developability properties, such as membrane permeability. Vilya-1 demonstrates substantial improvements in geometric accuracy compared to physics-based methods, co-folding networks, and other deep-learning conformer generators. Beyond prediction, it supports generative applications, allowing for the design of new macrocycles with specific chemical, structural, and property profiles. This establishes Vilya-1 as a foundational tool to accelerate the development of next-generation macrocycle therapeutics.

Why it matters

Professionals in pharmaceutical research and drug discovery can leverage Vilya-1 to significantly accelerate the design and optimization of macrocyclic peptide therapeutics, reducing development time and costs.

How to implement this in your domain

  1. 1Explore integrating Vilya-1 into existing drug discovery pipelines for macrocycle design.
  2. 2Utilize Vilya-1 for rapid screening and prediction of macrocycle conformations and properties.
  3. 3Apply its generative capabilities to design novel macrocycles with desired therapeutic profiles.
  4. 4Validate Vilya-1's predictions against experimental data to refine and optimize its use in specific projects.
  5. 5Collaborate with computational chemists to fully leverage the model's all-atom representation for detailed structural analysis.

Original post by Vilya Research, :, Pascal Sturmfels, Milad Salem, Naozumi Hiranuma, Stephen Rettie, Xiaoliang Pan, Benjamin D. Sellers, Adam P. Moyer, Patrick J. Salveson, Ivan Anishchanka

"arXiv:2607.09998v1 Announce Type: new Abstract: Macrocyclic peptides are an increasingly important therapeutic modality, but existing computational methods for modeling their structures and properties are limited in scope and do not generalize well across the synthetically access…"

View on X

Originally posted by Vilya Research, :, Pascal Sturmfels, Milad Salem, Naozumi Hiranuma, Stephen Rettie, Xiaoliang Pan, Benjamin D. Sellers, Adam P. Moyer, Patrick J. Salveson, Ivan Anishchanka on X · view source

Want to go deeper?

Turn these trends into skills with Learnijoy's hands-on AI & tech courses.

Explore courses

More in AI Research

AI ResearchAI Engineering & DevTools

Emotional Preferences Regulate Goal Priorities in Reinforcement Learning Agents

This paper proposes a computational framework where higher-level goals autonomously generate state-dependent emotional preferences to regulate the priorities of competing lower-level objectives in reinforcement learning agents. It demonstrates how this emergent preference function exhibits contextual priority switching and improves performance over fixed-preference strategies in multi-objective exploration environments.

Shiqi Liu, Yihua Tan, Hu Fu, Guanyu QiAug 28, 2026
AI Engineering & DevToolsAI Research

New Framework Unifies Task Detection and Adaptation for Continual Learning

This paper proposes FiUni, a Fisher-guided unified framework for task-free continual learning in LLMs that combines batch-level task detection with parameter-efficient adaptation. FiUni uses Fisher information matrix (FIM) properties to dynamically determine whether to reuse, expand, or create new low-rank adaptation (LoRA) subspaces, effectively mitigating catastrophic forgetting without explicit task boundaries.

Dezheng Han, Anbang Zhang, Zhihao Zhu, Shuaishuai GuoAug 28, 2026
AI Engineering & DevToolsAI Research

Soft EMG Interface Enables Machine Learning-Powered Silent Speech Recognition

This paper introduces a soft, active electromyography (EMG) interface worn on the hand that enables word-level silent speech recognition (SSR) using machine learning. The device acquires stable EMG signals from a fingertip electrode near the lips, achieving 97.2% accuracy on a 30-word vocabulary and demonstrating real-time drone control in noisy environments.

Yuta Kurotaki, Shusuke Yamakoshi, Reitaro Yoshida, Yutaka Isoda, Tamami Takano, Yuji Isano, Yusuke Miyake, Kentaro Kuribayashi, Hiroki OtaAug 28, 2026