Look Up table with Embeddings - UniProt - 2025 - Non Computational annotations

This PostgreSQL database contains structured information extracted from the UniProt API (retrieved in April 2025). It includes: 126,582 proteins 123,518 sequences 494.072 embeddings generated with ProtT5, ProSTT5, ESM2, Ankh 623,134 GO term annotations (with evidence codes: EXP, IDA, IPI, IMP, IGI, IEP, TAS, IC) Associated biological metadata The data was extracted using the protein-information-system repository and is used within the FANTASIA pipeline for automated functional annotation of protein sequences. The following UniProt API filter was applied to retrieve annotations:https://www.uniprot.org/uniprotkb?query=%28+go_exp%3A*+OR+go_ida%3A*+OR+go_ipi%3A*+OR+go_imp%3A*+OR+go_igi%3A*+OR+go_iep%3A*+OR+go_tas%3A*+OR+go_ic%3A*%29To ensure embedding reproducibility, all sequences were processed with batch size = 1, avoiding discrepancies caused by padding artifacts common in PLMs like ProtT5. To initialize the database, either of the following methods can be used: Option 1: Using pg_restore pg_restore -U usuario -h localhost -p 5432 -d BioData ./BioData_backup_2025_hq.tar Option 2: Using the FANTASIA CLI fantasia initialize --embeddings_url https://zenodo.org/records/15704357/files/PIS_2025_ankh_exp.tar?download=1 Notes: