S2DA-GO, a sequence-based model for Gene Ontology protein function prediction, is published in Frontiers in Genetics. Dong Wang is a corresponding author.
The method pairs protein language-model embeddings with semantic priors from GO term text, and uses a gradient-decoupled cross-attention branch so sparsely annotated functions interfere less with the shared sequence representation.
- Paper: doi:10.3389/fgene.2026.1880976
- Code: geneticodelab/S2DA-GO
Preferred citation, as given in the code repository:
Wang H, Xiang F, Zhang J, Wu T, Wang D, Wang Q, Lu B and Fu F (2026) S2DA-GO: enhancing protein function prediction via gradient-decoupled cross-attention and semantic priors. Front. Genet. 17:1880976. doi: 10.3389/fgene.2026.1880976