SciGroveBeta
Cheminformatics

Vilya-1: An all-atom foundation model for macrocycle structure prediction and design

Pascal Sturmfels, Milad Salem, Naozumi Hiranuma, Stephen Rettie, Xiaoliang Pan, Benjamin D. Sellers, Adam P. Moyer, Patrick J. Salveson, Ivan Anishchanka

Featured July 23, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A new AI model called Vilya-1 learns to predict the 3D shapes of complex ring-shaped molecules, called macrocycles, by looking at every single atom, helping scientists design new medicines faster.

In depth
The paper introduces Vilya-1, a deep learning model that uses an all-atom representation to predict macrocycle structures and properties. Unlike previous methods limited to specific chemistries, Vilya-1's unified architecture generalizes across diverse macrocycles and even small molecules, enabling accurate conformer generation and downstream tasks like property prediction and generative design.

Key Takeaways

  • 1
    Vilya-1 employs a uniform all-atom representation that allows it to model diverse macrocycle chemistries (peptidic, non-peptidic, canonical, non-canonical) and small molecules within a single architecture.
  • 2
    The model significantly improves geometric accuracy in sampling biologically relevant macrocycle conformations, outperforming physics-based methods, co-folding networks, and other deep-learning conformer generators.
  • 3
    Beyond structure prediction, Vilya-1 functions as a foundation model, fine-tunable for confidence estimation, property prediction (e.g., permeability), and generative design workflows, accelerating drug discovery.

Conceptual Flow

HIGH LEVEL
1
Methodology: Learning All Molecular Shapes

The computer learns to predict all possible 3D shapes of molecules by looking at every tiny atom, not just big parts, making it super flexible for different kinds of molecules.

Atom Details
Bond Details
Learn Patterns
Possible 3D Shapes
2
Results: Better Predictions, Faster Design

This new method is much better at guessing the correct molecule shapes than old ways, which helps scientists find new medicines more quickly and accurately.

Old Way
New Way
Compare Accuracy
Much Better Shapes
Faster Drug Design