FANTASIA

FANTASIA integrates protein language models for large-scale functional annotation of proteins via GO term transference using embedding space similarity.

Key Features:

• Protein language model integration: FANTASIA incorporates multiple protein language models including ProtTrans for per-protein annotation.
• Embedding space similarity: The tool transfers GO terms based on similarity in protein embedding space rather than traditional sequence homology.
• Enhanced model support: Latest developments include additional protein language models beyond the original ProtTrans implementation.
• High-throughput implementation: FANTASIA enables large-scale annotation of uncharacterized proteomes across diverse taxa.

Scientific Applications:

• Functional dark proteome annotation: FANTASIA predicts functions for virtually all proteins in proteomes, including up to 50% that remain unannotated by traditional homology-based methods.
• Non-model organism functional discovery: The tool offers advantages in sensitivity and generalizability for functional annotation in non-model taxa.
• Transcriptomics enrichment analysis: FANTASIA outputs support enrichment analyses from transcriptomics data.
• Novel gene function assignment: The tool assigns novel functions to unannotated genes in model organisms.
• Functionally important gene identification: FANTASIA identifies genes involved in important functions in non-model organisms.
• Molecular evolution and organismal biology: Large-scale functional predictions enhance understanding of molecular evolution across the animal tree of life.

Methodology:

FANTASIA computes protein embeddings using protein language models and transfers GO term annotations based on embedding space similarity.