Guides

Pre-built indexes

Published Centrifuger indexes for RefSeq, GTDB, NCBI core nt and NCBI nr, and how to download them.

Building a comprehensive index is an hours-to-days job on a large machine. For most projects there is no need: the Centrifuger authors publish ready-made indexes for the databases people ask for most often. Download one, point -x at it, and start classifying.

Available indexes#

The Size/~Memory column is both the download size and a good estimate of the RAM needed to classify against the index.

Title	Description	Link	Size/~Memory	Date
Genome databases
`cfr_hpv+gbsarscov2`	RefSeq human, bacteria, archaea, virus + SARS-CoV-2 variants from GenBank	Zenodo	41G	2023/10/01
`cfr_gtdb_r232`	GTDB r232	Dropbox	230G	2026/05/15
`cfr_gtdb_r232+refseq_hvfpc`	GTDB r232 + RefSeq human, virus, fungi, protozoa and contaminants (UniVec, EmVec)	Dropbox	232G	2026/05/15
`cfr_core_nt`	NCBI core nt	Dropbox	212G	2025/06/11
`cfr_llnl_core_nt_202603`	LLNL-curated NCBI core_nt (PMID:40111052)	Dropbox	242G	2026/03/01
`cfr_llnl_core_nt_wseqid_202512`	LLNL-curated NCBI core_nt with sequence ID information but less synchronised taxdmp	Dropbox	301G	2025/12/01
Protein databases
`cfr_protein_pv`	RefSeq bacteria, archaea and virus proteins	Zenodo	25G	2025/08/25
`cfr_nr`	NCBI nr	Dropbox	181G	2026/01/31

Older indexes remain available in this Dropbox folder.

Downloading with centrifuger-download#

The simplest route is to let Centrifuger fetch the index by title:

bash

centrifuger-download cfr_hpv+gbsarscov2

Substitute any title from the first column of the table.

Downloading by hand#

For the Dropbox-hosted indexes you may prefer to pull the files yourself — for instance on a cluster node where you want to control where the download lands, or to resume a partial transfer. Open the Dropbox folder in a browser, right-click each file and choose Copy link, then fetch it with wget:

bash

wget -O cfr_core_nt.1.cfr "https://www.dropbox.com/...&dl=1"

Download every *.cfr file belonging to the index and keep them together in one directory with a common prefix.

Choosing an index#

If you want to…	Consider
Classify human-associated microbiome samples on a mid-sized server	`cfr_hpv+gbsarscov2` (41G)
Use a standardised, rank-normalised bacterial and archaeal taxonomy	`cfr_gtdb_r232`
Cover prokaryotes and host, fungi, protozoa and vector contaminants	`cfr_gtdb_r232+refseq_hvfpc`
Search the broadest nucleotide collection, including eukaryotes	`cfr_core_nt` or an LLNL-curated variant
Detect divergent organisms through translated search	`cfr_protein_pv` or `cfr_nr`

Two practical considerations tend to decide it:

Memory. The index must fit in RAM. A 230 GB index needs a machine with well over 230 GB — the 41 GB RefSeq index runs comfortably where the large ones cannot run at all.
Host reads. If your sample contains host DNA, prefer an index that includes the host genome. Reads that have nowhere correct to go are the main source of spurious microbial calls.

Using a downloaded index#

Pass the prefix — the path without the .1.cfr suffix — to -x:

bash

centrifuger -x /data/indexes/cfr_hpv+gbsarscov2 \
  -1 sample_1.fq.gz -2 sample_2.fq.gz -t 16 > classification.tsv

centrifuger-quant -x /data/indexes/cfr_hpv+gbsarscov2 \
  -c classification.tsv > report.tsv

If none of these databases fits your project, build your own.

Edit this page on GitHub