sylph databases for the MGnify genome catalogues (19 biomes)

Pre-built sylph databases built from the MGnify genome catalogues, one file per biome. Ready for sylph profile on the command line, and used by PeekMicrobiome in the browser. Built with sylph 0.9.0, sylph sketch -c 200 -k 31 (default --min-spacing 30, pseudotax tracking enabled). The version matters: the .syldb format has changed across sylph releases, and a database has to be read by a version that understands it. Re-sketch from the source genomes if you need another c or k — the pipeline that produced these is scripts/build_biome_dbs.sh in the PeekMicrobiome repository. 19 catalogues, 56,782 species representatives. Each database was sketched from the species representatives of its catalogue and verified with sylph inspect before publication; the species counts below are the ones sylph reports, not the ones the API announces. They are not interchangeable and must not be merged. MGnify dereplicates each catalogue independently and they overlap — 10% of the named species are shared between human-oral and human-skin. Loading two at once lists the same species twice and splits its k-mers arbitrarily: sylph's pseudotax reassignment settles close genomes within one database, it does not arbitrate between two separately dereplicated catalogues. Profile against the catalogue matching your sample's biome. catalogue MGnify version species size soil v1.0 19,472 2822 MB marine v2.0 13,223 1222 MB marine-sediment v1.0 6,158 674 MB human-gut v2.0.2 4,744 433 MB mouse-gut v1.0 2,847 274 MB cow-rumen v1.0.1 2,729 237 MB sheep-rumen v1.0 2,172 179 MB chicken-gut v1.0.1 1,322 115 MB pig-gut v1.0 1,376 108 MB tomato-rhizosphere v1.0 579 91 MB human-skin v1.0 579 59 MB maize-rhizosphere v1.0 336 53 MB human-oral v1.0.1 452 32 MB human-vaginal v1.0 280 18 MB honeybee-gut v1.0.1 154 16 MB non-model-fish-gut v2.0 178 15 MB barley-rhizosphere v2.0 86 15 MB marine-eukaryotes vbeta 16 13 MB zebrafish-fecal v1.0 79 12 MB Please cite, if you use these: MGnify, the catalogue paper for the biome you profiled against, and sylph (Shaw & Yu, Nat. Biotechnol. 2024). Released under CC0: no rights are asserted over the sketches themselves. A .syldb is a FracMinHash of genomes that are not ours — a mechanical transform, with no creative step to own. EMBL-EBI places no restriction on redistribution and asks for attribution as good scientific practice, which is what the citations above are for. The databases are data, not code: the GPL of PeekMicrobiome does not cover them.