SPACE: STRING proteins as complementary embeddings

This repository holds the dataset used and generated in the research: SPACE: STRING proteins as complementary embeddings (at Bioinformatics). For reimplementation, please refer to the GitHub repository: https://github.com/deweihu96/SPACE Data descriptions: benchmarks.zip: benchmark datasets, including the subcellular localization dataset from DeepLoc 2.0 and the function prediction dataset from NetGO 2.0; please cite the corresponding papers if you use them: Thumuluri, Vineet, et al. "DeepLoc 2.0: multi-label subcellular localization prediction using protein language models." Nucleic acids research 50.W1 (2022): W228-W234. Yao, Shuwei, et al. "NetGO 2.0: improving large-scale protein function prediction with massive sequence, text, domain, family and network information." Nucleic acids research 49.W1 (2021): W469-W475. benchmark_results.zip: benchmark results. eggnog.zip: ortholog groups organized per taxonomy level, sourced from eggNOG 6.0; please cite the eggNOG database if you use this dataset: Hernández-Plaza, Ana, et al. "eggNOG 6.0: enabling comparative genomics across 12 535 organisms." Nucleic Acids Research 51.D1 (2023): D389-D394. euk_seed_groups.json: seed species kingdoms. euks_ancestors.tsv: species pairs and their latest common ancestor. euks.txt: all eukaryotic species used in the project. functional_embeddings.zip: the aligned, cross-species functional network embeddings, 512 dimensions. Files are per species. node2vec.zip: single species network embeddings generated by node2vec, 128 dimensions. The aligned functional embeddings and ProtT5 embeddings are also available from the STRING database.