Structure-Aware Masking for Protein Representation Learning
For researchers working on protein representation learning, this work provides a simple yet effective inductive bias that improves prediction of functional effects, particularly for higher-order mutations.
The paper introduces Bucket Masking, a structure-aware masking strategy for protein language models that masks groups of residues based on their 3D proximity. It achieves up to a 14% improvement over standard random masking on four protein fitness prediction tasks.
Masked language modeling (MLM) is the standard objective for training protein language models, typically implemented by randomly masking individual residues at a fixed rate (e.g., 15%). This practice implicitly assumes that all sequence positions contribute equally to representation learning. In downstream fitness prediction tasks, however, protein sequences are governed by three-dimensional structural dependencies and long-range residue contacts that induce strong nonlocal couplings between residues. We introduce Bucket Masking, a structure-aware masking strategy that selects groups of residues based on their proximity in three-dimensional space, preferentially masking structurally coupled regions during training. By conditioning the masking distribution on residue contacts, Bucket Masking shifts the learning objective toward modeling long-range interactions that are critical for protein function. Across four downstream protein fitness prediction tasks, Bucket Masking enables up to a 14% improvement over standard random masking, excelling at predicting higher-order mutational interactions. Through controlled ablations, we show that these improvements arise from mask placement rather than span size, establishing masking as a positional inductive bias.