Datasets & Resources

Human reference proteins are derived from the complete protein FASTA for NCBI RefSeq assembly GCF_000001405.40 (GRCh38.p14). Viral reference proteins are obtained from all UniProtKB entries matching taxonomy_id:10239, including reviewed and unreviewed entries and descendant taxa. Within each library, identical complete sequences are merged and all original accessions are retained. Candidate IDs identify sequences in this reference-library version; RefSeq and UniProt accessions are reported separately.

Reference libraries

Human90,373sequences

NCBI RefSeq human proteins: GCF_000001405.40 (GRCh38.p14)

Reference version

human_53e4a96179d0bb928cd9207b34689c01c2ee6548807d94b43ad064bbd8e90245

Source dataset version

30c771fef9c89edcced034e9065040618f5c468e5d9db54c6c3800f4b264e132

Viral866,479sequences

UniProtKB viral proteins: taxonomy_id:10239

Reference version

viral_258cdc8fbfcd07ac50d281f6dd68164951120999a4df558bccbc07ecc05ad7e8

Source dataset version

2b7fdffaf3e79107416d474e0a9f18fd000d6e1673cb6941f048149b077a24f8

Human and viral proteins are normalized and deduplicated by complete sequence. Their embeddings are computed with the fixed ESM-2 encoder and saved with the reference-library version.

Coverage is limited to the recorded source versions. Natural occurrence and embedding similarity do not establish an interaction experimentally.

Sources and applicable terms

Reference proteins follow the applicable NCBI and UniProt terms. UniProt copyrightable database content uses CC BY 4.0. The classification training records originate from HVIDB, which states no unified compilation license. Preserve the attribution and terms of each source.

The data and licensing documents are reproduced from the core source version used by this website. Original wording and attribution are retained. These documents do not grant additional rights to third-party data.