Struct2GO-Enhanced: Multimodal Graph Attention Improves Protein Function Prediction.

More recently, protein language models such as ESM-2 and ProtT5 have further advanced sequence-based annotation by learning high-capacity representations directly from large-scale unaligned protein corpora, highlighting an important trend toward integrating protein-language-model-derived features with structural models. , The emergence of AlphaFold2 has inaugurated a new era in structure-based function prediction, reinforcing the principle that structure determines function. , Studies such as DeepFRI leveraged experimentally determined structural databases for annotation, while Struct2GO and StructSeq2GO combined AlphaFold2 predictions with GNNs to improve accuracy.

also in 592 other papers