Unlocking Protein Dynamics: BioEmu is Now Available on Vecura

What is BioEmu?

BioEmu (Biomolecular Emulator) is a generative diffusion model designed to approximate the equilibrium Boltzmann distribution of protein structures directly from an amino acid sequence. Unlike traditional methods that focus on predicting a single lowest-energy structure, BioEmu samples the full conformational ensemble, including both folded states and flexible, disordered regions. It combines inlined ColabFold/AlphaFold2 MSA embeddings with a score-based denoiser trained on large-scale molecular dynamics (MD) data, allowing it to emulate microsecond-scale MD simulations with high efficiency.

What can users do with BioEmu on Vecura?

With BioEmu on Vecura, users can:

• Generate an ensemble of protein backbone conformations directly from a single sequence in minutes.
• Characterize conformational flexibility and identify potentially cryptic binding pockets.
• Perform side-chain reconstruction to generate all-atom models suitable for downstream structural analysis.
• Execute automated MD equilibration protocols to resolve steric clashes and refine structural models for docking or drug discovery workflows.

What the output means

The output provides a comprehensive structural ensemble—typically represented as a PDB topology file and an XTC trajectory file. These files encapsulate the range of physical conformations the protein is likely to adopt at equilibrium.

This output should be used to support scientific decision-making, such as guiding structural biology research or drug design hypotheses. It serves as a computational surrogate and does not replace experimental validation, such as NMR or X-ray crystallography.

Why this matters

Understanding how proteins behave at equilibrium—including their dynamic fluctuations—is critical for grasping their functional mechanisms. Traditional all-atom molecular dynamics simulations are computationally expensive, often limiting their use to short timescales or smaller systems.

BioEmu democratizes access to ensemble-level structural data by reducing the computational barrier to entry. By providing a scalable way to model protein dynamics, it enables researchers to conduct large-scale screenings and study flexible regions that were previously too time-consuming to simulate, accelerating discoveries in protein engineering and therapeutic design.

• Developed by: Microsoft Research
• Source: Official GitHub repository
• Reference: Science 2025 paper — Scalable emulation of protein equilibrium ensembles with generative deep learning