SLAE: Strictly Local All-atom Environment for Protein Representation

Building physically grounded protein representations is central to computational biology, yet most existing approaches rely on sequence-pretrained language models or backbone-only graphs that overlook side-chain geometry and chemical detail. We present SLAE, a unified all-atom framework for learning protein representations from each residue’s local atomic neighborhood using only atom types and interatomic geometries. To encourage expressive feature extraction, we introduce a novel multi-task autoencoder objective that combines coordinate reconstruction, sequence recovery, and energy regression. SLAE reconstructs allatom structures with high fidelity from latent residue environments and achieves state-of-the-art performance across diverse downstream tasks via transfer learning. SLAE’s latent space is chemically informative and environmentally sensitive, enabling quantitative assessment of structural qualities and smooth interpolation between conformations at all-atom resolution.

Proteins are tiny molecular machines whose atomic structures determine how they work, but capturing those structures in a useful form remains difficult. Many current methods describe proteins using only their letter sequence or a simplified protein backbone, which can leave out important atom-level chemical details. We developed SLAE, a method that learns from the full atomic neighborhood around each part of a protein. It compresses these local atomic environments into compact representations and learns to rebuild the complete 3D structure, including all atoms. During training, SLAE also learns to recover protein sequences and predict physical energy. SLAE learns representations that capture both backbone arrangement and atom-level chemistry. These representations improve a range of protein prediction tasks and can smoothly connect different conformations of the same protein at all-atom resolution. This helps move protein learning methods closer to understanding proteins as physical molecular structures, not just as strings of letters or simplified backbones.