================================================================================ Thal-Kak MSA databases Copyright 2026 CSSB, Seoul National University LICENSE AND ATTRIBUTION FOR THE DATABASES IN THIS BUCKET ================================================================================ This bucket distributes reference databases only. It contains no source code, and the Apache License 2.0 that covers the Thal-Kak source code (https://github.com/CSSB-SNU/Thal-Kak) does NOT apply to any file here. Each database is governed by the license named in its entry below, on the `Our license` line. The entries below are adapted from the reference-database section of the Thal-Kak NOTICE file and cover only the databases in this bucket. Which entry covers which archive in this bucket: 1. UniProt UniRef100 ........ hhblits/uniref100_2026_01.tar.zst 2. BFD, reduced subset ...... mmseqs/bfd_reduced.tar.zst 3. MGnify ................... mmseqs/mgnify_clusters.tar.zst hhblits/mgnify_clusters.tar.zst 4. EnVhogDB ................. mmseqs/envhog_std.tar.zst hhblits/envhog.tar.zst 5. Logan .................... mmseqs/logan_human.tar.zst hhblits/logan_human.tar.zst hhblits/logan_nonhuman.tar.zst ================================================================================ Reference databases in this bucket ================================================================================ The databases in this bucket are published by us and are DERIVED databases: each was reformatted into an MMseqs2 or HH-suite database, and some were re-clustered. They are NOT the original upstream distributions, and figures computed on them will not match figures computed on the upstream files. WHAT LICENSE APPLIES TO WHAT WE PUBLISH. Every database below derives from a single upstream source, so no database mixes licenses. For each source we name the upstream license as the distributor we obtained the data from states it, and then, separately, the license we place on the database we publish, under `Our license`. That second field is the one that governs what you may do with the files you downloaded from us. Where the upstream license is disputed between authorities, the dispute is written out and we follow the more restrictive claim for anything we redistribute. Of the HH-suite databases here, all but one were converted here from the corresponding MMseqs2 database, with their cs219 and hhm profiles computed here. For hhblits/uniref100_2026_01 and hhblits/logan_nonhuman, that MMseqs2 database is not in this bucket. The one exception is hhblits/envhog, which is upstream's own HH-suite build; its entry says so. Several of these databases also cover fewer clusters than the MMseqs2 database they were built from; each entry's `Changes` field says whether entries were excluded. One thing this notice does NOT do is license the output of a search. An alignment produced by a search contains material taken from whichever databases you searched, so the Apache-2.0 license on the Thal-Kak source code does not carry over to it - and we make no claim over it either. Everything below is provided AS IS, WITHOUT WARRANTY OF ANY KIND, either by the upstream provider or by us, and without any representation that it is fit for a particular purpose. -------------------------------------------------------------------------------- 1. UniProt UniRef100 Archive hhblits/uniref100_2026_01.tar.zst Upstream UniProt UniRef100, release 2026_01 Obtained https://ftp.uniprot.org/pub/databases/uniprot/uniref/ uniref100/uniref100.fasta.gz Source https://www.uniprot.org/ Accessed 2026-04-08 (the transfer ran into 2026-04-09) License CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ (https://www.uniprot.org/help/license, accessed 2026-08-13) Our license CC BY 4.0 Attribution The UniProt Consortium Cite The UniProt Consortium. UniProt: the Universal Protein Knowledgebase in 2025. Nucleic Acids Res 53(D1), D609-D617 (2025). doi:10.1093/nar/gkae1010 Changes The upstream sequences were RE-CLUSTERED into an MMseqs2 expandable-profile database, with every upstream sequence retained as a cluster member, and this HH-suite database was converted here from it. No entries were excluded. 2. BFD, reduced subset ("small BFD") Archive mmseqs/bfd_reduced.tar.zst Upstream BFD (Big Fantastic Database) by Steinegger M. and Soeding J.; the reduced subset - the first non-consensus sequence of each BFD cluster - was derived from it by Google DeepMind and is distributed as part of the AlphaFold genetic databases Obtained https://storage.googleapis.com/alphafold-databases/reduced_dbs/ bfd-first_non_consensus_sequences.fasta.gz Accessed 2025-02-04 License CC BY-SA 4.0 https://creativecommons.org/licenses/by-sa/4.0/ This is the notice its distributor attaches to the file we obtained: the AlphaFold README, under "Third-party database licenses", for "BFD (modified) ... modified by DeepMind" (accessed 2026-08-13). Our license CC BY-SA 4.0 https://creativecommons.org/licenses/by-sa/4.0/ We are passing along the license we received, not tightening one. Recipients of the underlying BFD sequences may also obtain them from their authors under CC BY 4.0 - see the Note. Attribution Martin Steinegger and Johannes Soeding for BFD; Google DeepMind for the reduced subset Cite Steinegger M. & Soeding J. Clustering huge protein sequence sets in linear time. Nat Commun 9, 2542 (2018). doi:10.1038/s41467-018-04964-5 For the reduced subset: Jumper J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583-589 (2021). doi:10.1038/s41586-021-03819-2 Changes Built here into an MMseqs2 expandable-profile database, which included re-clustering the sequences. `.idx` not included. Note The full BFD in HH-suite format is not in this bucket. The original authors distribute BFD themselves, under CC BY 4.0: "All files are available under a Creative Commons Attribution 4.0 International License" (https://bfd.mmseqs.com/, accessed 2026-08-13). The AlphaFold 3 README likewise states CC BY 4.0 for BFD (modified). Our CC BY-SA 4.0 label follows the notice on the copy we took and does not purport to withdraw anyone's CC BY 4.0 rights in the underlying BFD. 3. MGnify protein database Archives mmseqs/mgnify_clusters.tar.zst, hhblits/mgnify_clusters.tar.zst Upstream MGnify protein database, release 2024_04 Source https://www.ebi.ac.uk/metagenomics/ Accessed 2026-03-31 License CC0 1.0 https://creativecommons.org/publicdomain/zero/1.0/ (as stated by EMBL-EBI; the same terms are recorded in the AlphaFold and AlphaFold 3 database-license notices, accessed 2026-08-13) Our license CC0 1.0 for both databases. We assert no rights of our own. Attribution EMBL-EBI MGnify team Cite Richardson L. et al. MGnify: the microbiome sequence data analysis resource in 2023. Nucleic Acids Res 51(D1), D753-D759 (2023). doi:10.1093/nar/gkac1080 Changes mmseqs: the upstream sequences were RE-CLUSTERED into an MMseqs2 expandable-profile database. `.idx` not included. hhblits: converted here from that MMseqs2 database, and entries WERE excluded - clusters holding fewer than 3 sequences are absent, which is most of them. 4. EnVhogDB Archives mmseqs/envhog_std.tar.zst, hhblits/envhog.tar.zst Upstream EnVhogDB - Environmental Viral Homologous Groups of proteins Obtained http://envhog.u-ga.fr/envhog/ Accessed 2026-05-13 License CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ The download page we obtained the data from carries NO license statement of any kind. The same authors deposit the same dataset on Zenodo under CC BY 4.0, and we rely on that grant: 10.5281/zenodo.17086476 EnVhogDB raw data https://zenodo.org/records/17086059 EnVhogDB profiles both published 2025-09-09, both accessed 2026-08-13. Our license CC BY 4.0 for both databases. Attribution Ruben Perez-Bucio, Francois Enault and Clovis Galiez (Universite Grenoble Alpes) Cite Perez-Bucio R., Enault F., Galiez C. EnVhogDB: an extended view of the viral protein families on Earth through a vast collection of HMM profiles. Peer Community Journal, Volume 5 (2025), article no. e100. doi:10.24072/pcjournal.627 (The download page still cites the earlier preprint, bioRxiv 2024.06.25.600602.) Changes mmseqs/envhog_std: built here from the upstream protein FASTA and flat cluster TSV, re-clustered at the upstream `standard` level. `.idx` not included. hhblits/envhog: unlike the other HH-suite databases here, this one is NOT built from an MMseqs2 database - it is upstream's own HH-suite build, obtained as `enVhog_hhblits.tar.gz`. Its a3m and cs219 files are passed through unmodified. The only change is that hhm profiles for shallow alignments were removed. 5. Logan Archives mmseqs/logan_human.tar.zst, hhblits/logan_human.tar.zst, hhblits/logan_nonhuman.tar.zst Upstream Logan 50%-identity protein catalogue (Logan v1; SRA cutoff 2023-12-10) Obtained s3://logan-pub/p/ Source https://github.com/IndexThePlanet/Logan https://registry.opendata.aws/pasteur-logan/ Accessed 2026-04-03 License No license instrument has been applied to this data by its distributor. Logan's own entry on the AWS Open Data registry, and ours for these archives, name "NCBI Policy and NIH Genomic Data Sharing Policy", which is a policy pointer, not a grant. The Logan authors write that they "release all Logan data into the public-domain" - reproduced here as a statement of their intent, not as a license. Our license NONE. We grant no rights in this data and make no representation about your right to use or redistribute it. We redistribute it subject to the pass-through notice below. Attribution the Logan project and its authors, Chikhi R. et al. (31 authors; see the citation below) Cite Chikhi R., Lemane T., Loll-Krippleber R., Montoliu-Nerin M., Raffestin B., et al. Logan: Planetary-Scale Genome Assembly Surveys Life's Diversity. bioRxiv 2024.07.30.605881. doi:10.1101/2024.07.30.605881 (a preprint as of 2026-08-13; cite the journal version if one has appeared) Changes mmseqs: built into an MMseqs2 expandable-profile database. `.idx` not included. hhblits: converted here from that MMseqs2 database, and entries WERE excluded from both Logan HH-suite databases - clusters holding fewer than 3 sequences are absent, which is the great majority. Naming "human" means the annotated organism of the source SRA accession is taxid 9606, not that the sequences are of human origin; upstream warns these accessions also carry contaminating viral, bacterial and other sequence. Note logan_nonhuman is published in HH-suite format but NOT in MMseqs2 format: the MMseqs2 build extracts to 1.4 TB and is not published. This asymmetry is deliberate. Upstream pass-through notice, reproduced as required by NCBI: Logan is derived from the public portion of the NCBI Sequence Read Archive. Per NCBI policy, NCBI places no restrictions on the use or distribution of the data contained in its molecular databases, and does not accept data whose submitter has requested restrictions on reuse or redistribution. However, some submitters of the original data (or the country of origin) may claim patent, copyright, or other intellectual property rights; because no rights are transferred from submitters to NCBI, NCBI has none to transfer onward, and neither NCBI nor we can grant unrestricted permission concerning use, copying, or distribution.