BioEmu: A New Deep Learning System for Protein Structures

• BioEmu, a deep learning system developed by Microsoft, Rice University, and Freie Universität, allows high-resolution protein flexibility modelling at scale.

• It predicts the full range of shapes a protein naturally explores under biological conditions, allowing for large-scale predictions of protein function.

• BioEmu is faster and cheaper, enabling large-scale predictions of protein function.

• It uses an AI diffusion model to train BioEmu, feeding it real protein structures, simulations, and mutant sequences.

• BioEmu excels at benchmarks, capturing large shape changes in enzymes, local unfolding, and cryptic pockets.

• It predicts 83% of large shifts and 70-81% of small changes accurately, including open and closed forms of a vital enzyme called adenylate kinase.

• However, BioEmu can’t show how a process unfolds, handling temperature shifts, membranes, cell walls, drug molecules, pH changes, and how proteins interact.

• As the system grows to handle more complex proteins and chemical interactions, researchers may still need experiments or older simulation methods to validate its proposals.

• BioEmu’s conceptual advance is clear, enabling large-scale drug discovery and function studies with fewer resource constraints.

• Future scientists will need a deep grounding in physics, chemistry, machine learning, and physical modelling to unlock the true potential of such hybrid approaches.