The human gut metagenome is the focus of a lot of research in our time. From this site you can download the HumGut genome collection and accompanying metadata. As described in our paper, genomes encountered in healthy human guts worldwide were ranked by prevalence and clustered by whole genome identity (97.5%). Genomes representing the clusters, 30 691 in total, were retained as HumGut.
If you use this resource, we would ask you to cite our paper:
The HumGut collection contains 30,691 genomes. In this compressed archive:
you find the compressed FASTA files, one for each genome. All files
have Header-lines equipped with the proper text for building a kraken2
or krakenUniq database, see sections below for more details on this.
This archive is roughly 18GB.
The tab-separated text file
lists metadata about each HumGut genome (30,691 rows). Below you find a description of all columns.
||Unique HumGut name for each genome.|
||Unique HumGut identifier for each genome. These are integers from 3 000 000 and up, a choice made to not interfere with the NCBI Taxonomy database integers.|
||The highest resolution cluster this genome belongs to (97.5% sequence identity).|
||The coarser resolution cluster this genome belongs to (95% sequence identity).|
||GTDB-tk genome taxonomy ID. Note that the GTDB database (https://gtdb.ecogenomic.org/) has no such integer identifiers, and we have just artificially created some here. This is required for building kraken2/bracken/krakenUniq databases using this taxonomy. These integers are from 4 000 000 and up, a choice made to not interfere with the NCBI Taxonomy or the HumGut_tax_id mentioned above.|
||Genome name as given by GTDB-tk.|
||The full GTDB-tk taxonomy, from domain and down.|
||The taxonomy identifier from the NCBI Taxonomy database (https://www.ncbi.nlm.nih.gov/taxonomy/).|
||Genome name at the NCBI Taxonomy database.|
||The rank at the NCBI Taxonomy database.|
||The average sequence identity across 3,534 healthy human gut metagenomes.|
||The number of metagenomes where the genome was found present, using ≥ 95% sequence identity as a threshold.|
||The estimated completeness (%) of the genome.|
||The estimated contamination (%) of the genome.|
||Genome GC content.|
||Number of base pairs in genome.|
||Either RefSeq (https://ftp.ncbi.nlm.nih.gov/genomes/refseq/) or UHGG (https://www.ebi.ac.uk/metagenomics/)|
||The completion level as listed in RefSeq, or MAG (all UHGG genomes).|
||Number of genomes in the same cluster of the highest resolution (97.5%).|
||Number of genomes in the same coarse cluster (95%).|
||The name of the FASTA file in the archive HumGut.tar.gz from above.|
||The ftp address from where we downloaded the genome.|
The tab-separated text file
lists metadata about all the 381,779 genomes used for obtaining the
HumGut clusters (381,779 rows). The columns are a subset of those in the
above table, see the above description of column names. Note that we do
not provide the FASTA files for all these genomes at this website, since
they are publicly available elsewhere. The FTP addresses in the column
ftp_download shows where each genome is found.
One obvious use of the HumGut collection is to assign some taxonomy
to the reads you have after sequencing a human gut metagenome. In the
table HumGut.tsv mentioned above, each HumGut genomes
has a taxonomy identifier (
HumGut_tax_id). You also find
the same table. These are the parents of the HumGut genome in
the GTDB or the NCBI taxonomy, respectively. Be aware that even if each
HumGut genome is clustered at a sub-species identity threshold, its
parent may not be a species, but sometimes a genus or even higher rank.
This may be because some branches in the taxonomy do not contain all
ranks, or the HumGut genome is simply too different from any species
listed by GTDB or NCBI, and has been assigned directly under some higher
In order to describe the taxonomy tree, we have chosen to use the data structures used by NCBI Taxonomy (https://ftp.ncbi.nih.gov/pub/taxonomy/). This is also what tools like kraken2 and krakenUniq uses (see below).
Here you can download the two files needed for using the GTDB taxonomy:
Here you can download the files needed for using the NCBI taxonomy:
Both these sets of files contain the
all HumGut genomes. Their parents are the corresponding
ncbi_tax_id. The files then
contain the branches leading down to these taxa. Note that the files
do not contain the full taxonomy for all taxa in either
database. They have been pruned to only contain the relevant branches
leading to some HumGut genome.
In addition, both pairs of files include the human genome branch, with its NCBI taxonomy. This has been included since we believe the human genome should always be included in a reference database for reads from the human gut (as possible contaminations).
Both taxonomies above are from January 2021. This changes slowly over
time. We will make efforts to update this at regular intervals.
The kraken2 software (https://github.com/DerrickWood/kraken2) is a popular tool for making taxonomic classification of metagenome reads. Building a kraken2 database from the HumGut genomes is an excellent tool for taxonomic profiling of data from the human gut. Below we describe a procedure for building the kraken2 database.
Make a folder in which you want to build the kraken2 database. We
refer to this as
Make the subfolder
$DBHOME/taxonomy. If you want to use
the GTDB taxonomy, copy
gtdb_nodes.dmp into this, or the
ncbi_nodes.dmp if you want
to use the NCBI taxonomy.
We strongly recommend you also include the human genome in the database as long as your data are from the human gut. Here is the code for including the human genome:
kraken2-build --download-library human --db $DBHOME
$DBHOME/library should appear, and inside it,
We assume you have extracted the
above into the folder
$fna If you include all HumGut
genomes, simply write all FASTA-files in this folder to a single
uncompressed FASTA file. The latter because kraken2 cannot build from
compressed FASTA files. It can be done like this
zcat $fna/*.fna.gz > HumGut975_library.fna
HumGut975_library.fna should be close to 60GB.
Then you add this to the kraken2 database with
kraken2-build --add-to-library HumGut975_library.fna --db $DBHOME
$DBHOME/library/added should now appear.You
could speed this up by using multiple threads (
kraken2-build). After this step, delete the huge
In the example above, we included all HumGut genomes in the database. If this requires too much memory, or you simply just want a lower resolution, you may only use the 95% clustered genomes in the database. Here is some R code for creating a corresponding library file:
library(tidyverse) read_delim("HumGut.tsv", delim = "\t") %>% distinct(cluster95, .keep_all = T) -> humgut.tbl ok <- file.append("HumGut95_library.fna.gz", file.path("fna", humgut.tbl$genome_file))
Note that in the above code we assume the file
HumGut.tsv and the folder
fna is in the
current working directory, please use correct paths if they are
elsewhere. Note also that in
HumGut.tsv the rows are sorted
in descending order by the
prevalence_score and hence, the
first row for each cluster is the genome to keep.
This code produces a compressed FASTA file, and you need to
uncompress it before you call
kraken2-build. This file
should be around 10GB when uncompressed.
You may of course also select all kinds of other subsets of the HumGut genomes to include in your database, using a similar approach.
The last step is just to build the database
kraken2-build --build --threads 20 --db $DBHOME
Here we used 20 threads. This step is the most time-consuming.