Original title: Accelerating Inference in Molecular Diffusion Models with Latent Representations of Protein Structure
Authors: Ian Dunn, David Ryan Koes
Within structural biology and drug design, diffusion generative models hold immense potential. They directly analyze 3D molecular structures, but their efficiency faces challenges due to graph neural networks’ scalability and slow inference speeds. To counter this, existing models simplify protein structure representations, sacrificing crucial interaction details and structural fidelity.
Original article: https://arxiv.org/abs/2311.13466